AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
feat(platform): split marketing into four scoped experts (#14543)
Ubbe14243e56a204
diff --git a/autogpt_platform/backend/agents/StoreAgent_rows.csv b/autogpt_platform/backend/agents/StoreAgent_rows.csvindex 44a5e052fc..96f775a01b 100644--- a/autogpt_platform/backend/agents/StoreAgent_rows.csv+++ b/autogpt_platform/backend/agents/StoreAgent_rows.csv@@ -240,3 +240,5 @@ Result: a list of dictionaries such as{ ""name"": ""Liam O’Connor"", ""date_of_birth"": ""1985‑07‑19"", ""job_title"": ""Marketing Manager"", ""age"": 39 },…18 more entries…]","[""development""]",false,true+d34ccb00-d9da-5466-a545-1aff5755513c,a1af1733-b566-5e2f-ae20-068e2163c4cc,lifecycle-email-sequence-builder,Lifecycle Email Sequence Builder,,,false,"Map the emails worth sending, then write them","Turn a product and an audience into a lifecycle email sequence that is mapped before it is written. The workflow first lays out which emails should exist at all — the moment each one attaches to, the trigger that fires it, its single goal and the one action it asks for — and only then drafts them. Works for welcome, onboarding, nurture and win-back sequences. Timing is anchored to behaviour rather than to a fixed calendar, every email has an exit, and anything the drafts cannot know is left as a marked gap rather than invented.","[""marketing""]",false,true+620abb2f-0592-50ff-beed-760e22fe735e,31d7abc3-e191-5cac-b3be-3e603fc49314,winback-email-writer,Win-back Email Writer,,,false,Bring quiet customers back without begging,"Write a win-back email to a customer or segment that has gone quiet, plus five subject lines and a three-touch follow-up plan that knows when to stop. The email acknowledges the gap without apologising, states the offer and its condition plainly, asks a question answerable in one line, and always gives an easy way out. No guilt, no false scarcity, and no invented purchase history — anything it has not been told is left as a marked gap.","[""marketing""]",false,truediff --git a/autogpt_platform/backend/agents/agent_31d7abc3-e191-5cac-b3be-3e603fc49314.json b/autogpt_platform/backend/agents/agent_31d7abc3-e191-5cac-b3be-3e603fc49314.jsonnew file mode 100644index 0000000000..5d09cc1160--- /dev/null+++ b/autogpt_platform/backend/agents/agent_31d7abc3-e191-5cac-b3be-3e603fc49314.json@@ -0,0 +1,465 @@+{+ "id": "964fa167-39d4-5d0f-83db-ba2571ce0880",+ "version": 1,+ "is_active": true,+ "name": "Win-back Email Writer",+ "description": "Write a win-back email to a customer or segment that has gone quiet, plus five subject lines and a three-touch follow-up plan that knows when to stop. The email acknowledges the gap without apologising, states the offer and its condition plainly, asks a question answerable in one line, and always gives an easy way out. No guilt, no false scarcity, and no invented purchase history \u2014 anything it has not been told is left as a marked gap.",+ "instructions": null,+ "recommended_schedule_cron": null,+ "nodes": [+ {+ "id": "2e02942b-4e6d-53ff-a0ad-4ab0e359e474",+ "block_id": "7fcd3bcb-8e1b-4e69-903d-32d3d4a92158",+ "input_default": {+ "name": "Contact or Segment",+ "secret": false,+ "advanced": false,+ "description": "Who you are writing to \u2014 a person, or the segment they belong to.",+ "placeholder_values": []+ },+ "metadata": {+ "position": {+ "x": -900.0,+ "y": 0.0+ }+ },+ "input_links": [],+ "output_links": [+ {+ "id": "270fc759-f1ef-5530-967f-fbbab3147416",+ "source_id": "2e02942b-4e6d-53ff-a0ad-4ab0e359e474",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_contact",+ "is_static": true+ }+ ],+ "graph_id": "964fa167-39d4-5d0f-83db-ba2571ce0880",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "4b8719f5-a5cd-5898-bf7e-223ee156fb87",+ "block_id": "90a56ffb-7024-4b2b-ab50-e26c5e5ab8ba",+ "input_default": {+ "name": "What Happened",+ "secret": false,+ "advanced": false,+ "description": "What they last did, when, and anything you know about why they went quiet.",+ "placeholder_values": []+ },+ "metadata": {+ "position": {+ "x": -900.0,+ "y": 320.0+ }+ },+ "input_links": [],+ "output_links": [+ {+ "id": "cfe190fd-728d-5970-8e28-437b18725b4c",+ "source_id": "4b8719f5-a5cd-5898-bf7e-223ee156fb87",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_history",+ "is_static": true+ }+ ],+ "graph_id": "964fa167-39d4-5d0f-83db-ba2571ce0880",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "bb749f13-cf59-5bf2-beba-7ad34a4b14fa",+ "block_id": "90a56ffb-7024-4b2b-ab50-e26c5e5ab8ba",+ "input_default": {+ "name": "The Offer",+ "secret": false,+ "advanced": false,+ "description": "What you can offer them now, including any condition attached to it.",+ "placeholder_values": []+ },+ "metadata": {+ "position": {+ "x": -900.0,+ "y": 640.0+ }+ },+ "input_links": [],+ "output_links": [+ {+ "id": "bd9cbcb1-d119-593c-bed4-a12c26d8b449",+ "source_id": "bb749f13-cf59-5bf2-beba-7ad34a4b14fa",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_offer",+ "is_static": true+ }+ ],+ "graph_id": "964fa167-39d4-5d0f-83db-ba2571ce0880",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "520d2fee-f12b-5099-9741-c133a1cfd8dc",+ "block_id": "7fcd3bcb-8e1b-4e69-903d-32d3d4a92158",+ "input_default": {+ "name": "Sender Name",+ "secret": false,+ "advanced": false,+ "description": "Who the email is signed by.",+ "placeholder_values": []+ },+ "metadata": {+ "position": {+ "x": -900.0,+ "y": 960.0+ }+ },+ "input_links": [],+ "output_links": [+ {+ "id": "78bf0755-d563-535a-a351-55d0e2566f4f",+ "source_id": "520d2fee-f12b-5099-9741-c133a1cfd8dc",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_sender",+ "is_static": true+ }+ ],+ "graph_id": "964fa167-39d4-5d0f-83db-ba2571ce0880",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "block_id": "1f292d4a-41a4-4977-9684-7c8d560b9f91",+ "input_default": {+ "model": "claude-sonnet-4-5-20250929",+ "retry": 3,+ "prompt": "You are writing a win-back email to someone who has gone quiet.\n\nWho they are: {{contact}}\nWhat happened, and why they may have gone quiet: {{history}}\nWhat we can offer them now: {{offer}}\nSigned by: {{sender}}\n\nReturn three things, in this order.\n\n## Five subject lines\n\nEach one under six words, lowercase, describing what is inside. Mark the one you\nwould send and say why in a single line.\n\n## The email\n\nSix sentences at most, in this order: what they did last and when, one honest\nsentence acknowledging the gap without apologising for existing, what has changed\nthat is relevant to them specifically, the offer stated plainly with its\ncondition, a question answerable in one line, and an easy way out.\n\nNo greeting beyond their first name. No \"we miss you\". No guilt, no false\nscarcity, no discount unless the offer given above contains one. Write it so the\nrecipient could forward it to a colleague without embarrassment.\n\n## The follow-up plan\n\nThree touches at most, then stop and mark them dormant. For each: how many days\nlater, what new thing it says, and the condition that cancels the rest. A reply\nof \"not now\" ends the sequence \u2014 record the reason and any date they suggest.\n\nNever invent a purchase history, a usage statistic, a previous conversation or a\nresult you cannot name the customer for. Where you need a fact you have not been\ngiven, leave a marked gap in square brackets and list what is missing at the\nend.",+ "sys_prompt": "You are a lifecycle email specialist. You write plainly, you never invent customer data, and you always leave the recipient an easy way out.",+ "prompt_values": {}+ },+ "metadata": {+ "position": {+ "x": 0.0,+ "y": 0.0+ }+ },+ "input_links": [+ {+ "id": "270fc759-f1ef-5530-967f-fbbab3147416",+ "source_id": "2e02942b-4e6d-53ff-a0ad-4ab0e359e474",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_contact",+ "is_static": true+ },+ {+ "id": "cfe190fd-728d-5970-8e28-437b18725b4c",+ "source_id": "4b8719f5-a5cd-5898-bf7e-223ee156fb87",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_history",+ "is_static": true+ },+ {+ "id": "bd9cbcb1-d119-593c-bed4-a12c26d8b449",+ "source_id": "bb749f13-cf59-5bf2-beba-7ad34a4b14fa",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_offer",+ "is_static": true+ },+ {+ "id": "78bf0755-d563-535a-a351-55d0e2566f4f",+ "source_id": "520d2fee-f12b-5099-9741-c133a1cfd8dc",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_sender",+ "is_static": true+ }+ ],+ "output_links": [+ {+ "id": "224f46f5-1308-5d8b-8e0e-0e21659de7a1",+ "source_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "sink_id": "ccd34d64-9fcf-540a-b576-923e14f1456a",+ "source_name": "response",+ "sink_name": "value",+ "is_static": false+ }+ ],+ "graph_id": "964fa167-39d4-5d0f-83db-ba2571ce0880",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "ccd34d64-9fcf-540a-b576-923e14f1456a",+ "block_id": "363ae599-353e-4804-937e-b2ee3cef3da4",+ "input_default": {+ "name": "Win-back Email",+ "secret": false,+ "advanced": false,+ "escape_html": false+ },+ "metadata": {+ "position": {+ "x": 900.0,+ "y": 0.0+ }+ },+ "input_links": [+ {+ "id": "224f46f5-1308-5d8b-8e0e-0e21659de7a1",+ "source_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "sink_id": "ccd34d64-9fcf-540a-b576-923e14f1456a",+ "source_name": "response",+ "sink_name": "value",+ "is_static": false+ }+ ],+ "output_links": [],+ "graph_id": "964fa167-39d4-5d0f-83db-ba2571ce0880",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ }+ ],+ "links": [+ {+ "id": "270fc759-f1ef-5530-967f-fbbab3147416",+ "source_id": "2e02942b-4e6d-53ff-a0ad-4ab0e359e474",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_contact",+ "is_static": true+ },+ {+ "id": "cfe190fd-728d-5970-8e28-437b18725b4c",+ "source_id": "4b8719f5-a5cd-5898-bf7e-223ee156fb87",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_history",+ "is_static": true+ },+ {+ "id": "bd9cbcb1-d119-593c-bed4-a12c26d8b449",+ "source_id": "bb749f13-cf59-5bf2-beba-7ad34a4b14fa",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_offer",+ "is_static": true+ },+ {+ "id": "78bf0755-d563-535a-a351-55d0e2566f4f",+ "source_id": "520d2fee-f12b-5099-9741-c133a1cfd8dc",+ "sink_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "source_name": "result",+ "sink_name": "prompt_values_#_sender",+ "is_static": true+ },+ {+ "id": "224f46f5-1308-5d8b-8e0e-0e21659de7a1",+ "source_id": "28a69a8e-8d38-5880-8cba-85257c524b17",+ "sink_id": "ccd34d64-9fcf-540a-b576-923e14f1456a",+ "source_name": "response",+ "sink_name": "value",+ "is_static": false+ }+ ],+ "forked_from_id": null,+ "forked_from_version": null,+ "sub_graphs": [],+ "user_id": "",+ "created_at": "2026-09-14T00:00:00.000Z",+ "input_schema": {+ "type": "object",+ "properties": {+ "Contact or Segment": {+ "advanced": false,+ "secret": false,+ "title": "Contact or Segment",+ "description": "Who you are writing to \u2014 a person, or the segment they belong to."+ },+ "What Happened": {+ "advanced": false,+ "secret": false,+ "title": "What Happened",+ "description": "What they last did, when, and anything you know about why they went quiet."+ },+ "The Offer": {+ "advanced": false,+ "secret": false,+ "title": "The Offer",+ "description": "What you can offer them now, including any condition attached to it."+ },+ "Sender Name": {+ "advanced": false,+ "secret": false,+ "title": "Sender Name",+ "description": "Who the email is signed by."+ }+ },+ "required": [+ "Contact or Segment",+ "What Happened",+ "The Offer",+ "Sender Name"+ ]+ },+ "output_schema": {+ "type": "object",+ "properties": {+ "Win-back Email": {+ "advanced": false,+ "secret": false,+ "title": "Win-back Email"+ }+ },+ "required": [+ "Win-back Email"+ ]+ },+ "has_external_trigger": false,+ "has_human_in_the_loop": false,+ "trigger_setup_info": null,+ "credentials_input_schema": {+ "properties": {+ "anthropic_api_key_credentials": {+ "credentials_provider": [+ "anthropic"+ ],+ "credentials_types": [+ "api_key"+ ],+ "properties": {+ "id": {+ "title": "Id",+ "type": "string"+ },+ "title": {+ "anyOf": [+ {+ "type": "string"+ },+ {+ "type": "null"+ }+ ],+ "default": null,+ "title": "Title"+ },+ "provider": {+ "const": "anthropic",+ "title": "Provider",+ "type": "string"+ },+ "type": {+ "const": "api_key",+ "title": "Type",+ "type": "string"+ }+ },+ "required": [+ "id",+ "provider",+ "type"+ ],+ "title": "CredentialsMetaInput[Literal[<ProviderName.ANTHROPIC: 'anthropic'>], Literal['api_key']]",+ "type": "object",+ "discriminator": "model",+ "discriminator_mapping": {+ "Llama-3.3-70B-Instruct": "llama_api",+ "Llama-3.3-8B-Instruct": "llama_api",+ "Llama-4-Maverick-17B-128E-Instruct-FP8": "llama_api",+ "Llama-4-Scout-17B-16E-Instruct-FP8": "llama_api",+ "Qwen/Qwen2.5-72B-Instruct-Turbo": "aiml_api",+ "amazon/nova-lite-v1": "open_router",+ "amazon/nova-micro-v1": "open_router",+ "amazon/nova-pro-v1": "open_router",+ "claude-3-7-sonnet-20250219": "anthropic",+ "claude-3-haiku-20240307": "anthropic",+ "claude-haiku-4-5-20251001": "anthropic",+ "claude-opus-4-1-20250805": "anthropic",+ "claude-opus-4-20250514": "anthropic",+ "claude-opus-4-5-20251101": "anthropic",+ "claude-sonnet-4-20250514": "anthropic",+ "claude-sonnet-4-5-20250929": "anthropic",+ "cohere/command-r-08-2024": "open_router",+ "cohere/command-r-plus-08-2024": "open_router",+ "deepseek/deepseek-chat": "open_router",+ "deepseek/deepseek-r1-0528": "open_router",+ "dolphin-mistral:latest": "ollama",+ "google/gemini-2.0-flash-001": "open_router",+ "google/gemini-2.0-flash-lite-001": "open_router",+ "google/gemini-2.5-flash": "open_router",+ "google/gemini-2.5-flash-lite-preview-06-17": "open_router",+ "google/gemini-2.5-pro-preview-03-25": "open_router",+ "google/gemini-3-pro-preview": "open_router",+ "gpt-3.5-turbo": "openai",+ "gpt-4-turbo": "openai",+ "gpt-4.1-2025-04-14": "openai",+ "gpt-4.1-mini-2025-04-14": "openai",+ "gpt-4o": "openai",+ "gpt-4o-mini": "openai",+ "gpt-5-2025-08-07": "openai",+ "gpt-5-chat-latest": "openai",+ "gpt-5-mini-2025-08-07": "openai",+ "gpt-5-nano-2025-08-07": "openai",+ "gpt-5.1-2025-11-13": "openai",+ "gryphe/mythomax-l2-13b": "open_router",+ "llama-3.1-8b-instant": "groq",+ "llama-3.3-70b-versatile": "groq",+ "llama3": "ollama",+ "llama3.1:405b": "ollama",+ "llama3.2": "ollama",+ "llama3.3": "ollama",+ "meta-llama/Llama-3.2-3B-Instruct-Turbo": "aiml_api",+ "meta-llama/Llama-3.3-70B-Instruct-Turbo": "aiml_api",+ "meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo": "aiml_api",+ "meta-llama/llama-4-maverick": "open_router",+ "meta-llama/llama-4-scout": "open_router",+ "microsoft/wizardlm-2-8x22b": "open_router",+ "mistralai/mistral-nemo": "open_router",+ "moonshotai/kimi-k2": "open_router",+ "nousresearch/hermes-3-llama-3.1-405b": "open_router",+ "nousresearch/hermes-3-llama-3.1-70b": "open_router",+ "nvidia/llama-3.1-nemotron-70b-instruct": "aiml_api",+ "o1": "openai",+ "o1-mini": "openai",+ "o3-2025-04-16": "openai",+ "o3-mini": "openai",+ "openai/gpt-oss-120b": "open_router",+ "openai/gpt-oss-20b": "open_router",+ "perplexity/sonar": "open_router",+ "perplexity/sonar-deep-research": "open_router",+ "perplexity/sonar-pro": "open_router",+ "qwen/qwen3-235b-a22b-thinking-2507": "open_router",+ "qwen/qwen3-coder": "open_router",+ "v0-1.0-md": "v0",+ "v0-1.5-lg": "v0",+ "v0-1.5-md": "v0",+ "x-ai/grok-4": "open_router",+ "x-ai/grok-4-fast": "open_router",+ "x-ai/grok-4.1-fast": "open_router",+ "x-ai/grok-code-fast-1": "open_router"+ },+ "discriminator_values": [+ "claude-sonnet-4-5-20250929"+ ]+ }+ },+ "required": [+ "anthropic_api_key_credentials"+ ],+ "title": "Credentials",+ "type": "object"+ }+}diff --git a/autogpt_platform/backend/agents/agent_a1af1733-b566-5e2f-ae20-068e2163c4cc.json b/autogpt_platform/backend/agents/agent_a1af1733-b566-5e2f-ae20-068e2163c4cc.jsonnew file mode 100644index 0000000000..e598272aec--- /dev/null+++ b/autogpt_platform/backend/agents/agent_a1af1733-b566-5e2f-ae20-068e2163c4cc.json@@ -0,0 +1,467 @@+{+ "id": "2ec2f82c-2928-5252-9ebf-184c98a22167",+ "version": 1,+ "is_active": true,+ "name": "Lifecycle Email Sequence Builder",+ "description": "Turn a product and an audience into a lifecycle email sequence that is mapped before it is written. The workflow first lays out which emails should exist at all \u2014 the moment each one attaches to, the trigger that fires it, its single goal and the one action it asks for \u2014 and only then drafts them. Works for welcome, onboarding, nurture and win-back sequences. Timing is anchored to behaviour rather than to a fixed calendar, every email has an exit, and anything the drafts cannot know is left as a marked gap rather than invented.",+ "instructions": null,+ "recommended_schedule_cron": null,+ "nodes": [+ {+ "id": "3bf3107e-ffa6-51fc-af23-3fff5737d602",+ "block_id": "7fcd3bcb-8e1b-4e69-903d-32d3d4a92158",+ "input_default": {+ "name": "Product",+ "secret": false,+ "advanced": false,+ "description": "What you sell, in a sentence.",+ "placeholder_values": []+ },+ "metadata": {+ "position": {+ "x": -900.0,+ "y": 0.0+ }+ },+ "input_links": [],+ "output_links": [+ {+ "id": "c5d67ddb-4d4d-59f9-8bc3-8a27f8cb0a84",+ "source_id": "3bf3107e-ffa6-51fc-af23-3fff5737d602",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_product",+ "is_static": true+ }+ ],+ "graph_id": "2ec2f82c-2928-5252-9ebf-184c98a22167",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "aa353f90-fdbc-5030-b5ac-3a9a2525d261",+ "block_id": "7fcd3bcb-8e1b-4e69-903d-32d3d4a92158",+ "input_default": {+ "name": "Audience",+ "secret": false,+ "advanced": false,+ "description": "Who receives these emails, and the situation they are in.",+ "placeholder_values": []+ },+ "metadata": {+ "position": {+ "x": -900.0,+ "y": 320.0+ }+ },+ "input_links": [],+ "output_links": [+ {+ "id": "4c09fda7-77a4-573f-953f-10153b96eef7",+ "source_id": "aa353f90-fdbc-5030-b5ac-3a9a2525d261",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_audience",+ "is_static": true+ }+ ],+ "graph_id": "2ec2f82c-2928-5252-9ebf-184c98a22167",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "b65bed88-e8b6-506a-b047-9ab5bb0d2a6f",+ "block_id": "7fcd3bcb-8e1b-4e69-903d-32d3d4a92158",+ "input_default": {+ "name": "Sequence Type",+ "value": "welcome",+ "secret": false,+ "advanced": false,+ "description": "welcome, onboarding, nurture, win-back, or re-engagement.",+ "placeholder_values": []+ },+ "metadata": {+ "position": {+ "x": -900.0,+ "y": 640.0+ }+ },+ "input_links": [],+ "output_links": [+ {+ "id": "d838804d-2354-51d9-b742-8481dd73a3c4",+ "source_id": "b65bed88-e8b6-506a-b047-9ab5bb0d2a6f",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_sequence_type",+ "is_static": true+ }+ ],+ "graph_id": "2ec2f82c-2928-5252-9ebf-184c98a22167",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "e816df80-90f7-5c9e-9612-26fd09247ba7",+ "block_id": "90a56ffb-7024-4b2b-ab50-e26c5e5ab8ba",+ "input_default": {+ "name": "Voice and Constraints",+ "value": "Clear and direct. No jargon, no exclamation marks, and never \"we're excited to announce\".",+ "secret": false,+ "advanced": false,+ "description": "How it should sound, plus anything it must not say or promise.",+ "placeholder_values": []+ },+ "metadata": {+ "position": {+ "x": -900.0,+ "y": 960.0+ }+ },+ "input_links": [],+ "output_links": [+ {+ "id": "b3924cc1-0b89-5f1d-a39e-7e297ea7de24",+ "source_id": "e816df80-90f7-5c9e-9612-26fd09247ba7",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_voice",+ "is_static": true+ }+ ],+ "graph_id": "2ec2f82c-2928-5252-9ebf-184c98a22167",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "block_id": "1f292d4a-41a4-4977-9684-7c8d560b9f91",+ "input_default": {+ "model": "claude-sonnet-4-5-20250929",+ "retry": 3,+ "prompt": "You are mapping and writing a lifecycle email sequence.\n\nProduct: {{product}}\nAudience: {{audience}}\nSequence type: {{sequence_type}}\nVoice and constraints: {{voice}}\n\nWork in two passes and show both.\n\n## Pass one \u2014 the map\n\nA table with one row per email: the moment it attaches to, the trigger that\nfires it, the single goal, the subject line, and the one action you want.\n\nAn email only earns a row if it attaches to something the reader did or failed\nto do. If a moment has no detectable trigger, leave it out and say so under the\ntable rather than filling the gap with a timed send.\n\nAnchor timing to behaviour, not to a fixed calendar. Never schedule more than\none automated email in any 48 hours. State explicitly which queued emails each\nbehavioural send cancels: only the ones that action makes redundant, never the\nwhole sequence.\n\n## Pass two \u2014 the drafts\n\nWrite every email in the map. For each one:\n\n- One goal, one link to it, and that link appears twice at most.\n- A subject line that describes what is inside. No curiosity gaps.\n- Plain text unless the content is genuinely visual.\n- Short paragraphs. No preamble before the point.\n- An exit: replying, or doing the thing being asked, ends the sequence.\n\nMatch the voice and constraints given above. If they are empty, write in clear\ndirect prose with the marketing jargon stripped out.\n\nNever invent product behaviour, customer numbers, testimonials or results. Where\na draft needs a fact you have not been given, leave a marked gap in square\nbrackets and list what is missing at the end.",+ "sys_prompt": "You are a lifecycle email specialist. You write plainly, you never invent customer data, and you always leave the recipient an easy way out.",+ "prompt_values": {}+ },+ "metadata": {+ "position": {+ "x": 0.0,+ "y": 0.0+ }+ },+ "input_links": [+ {+ "id": "c5d67ddb-4d4d-59f9-8bc3-8a27f8cb0a84",+ "source_id": "3bf3107e-ffa6-51fc-af23-3fff5737d602",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_product",+ "is_static": true+ },+ {+ "id": "4c09fda7-77a4-573f-953f-10153b96eef7",+ "source_id": "aa353f90-fdbc-5030-b5ac-3a9a2525d261",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_audience",+ "is_static": true+ },+ {+ "id": "d838804d-2354-51d9-b742-8481dd73a3c4",+ "source_id": "b65bed88-e8b6-506a-b047-9ab5bb0d2a6f",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_sequence_type",+ "is_static": true+ },+ {+ "id": "b3924cc1-0b89-5f1d-a39e-7e297ea7de24",+ "source_id": "e816df80-90f7-5c9e-9612-26fd09247ba7",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_voice",+ "is_static": true+ }+ ],+ "output_links": [+ {+ "id": "d94eec1c-4ebe-53b6-b321-fb194232ba50",+ "source_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "sink_id": "c6522752-9ff3-5d3f-96cf-4c38a89a9329",+ "source_name": "response",+ "sink_name": "value",+ "is_static": false+ }+ ],+ "graph_id": "2ec2f82c-2928-5252-9ebf-184c98a22167",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ },+ {+ "id": "c6522752-9ff3-5d3f-96cf-4c38a89a9329",+ "block_id": "363ae599-353e-4804-937e-b2ee3cef3da4",+ "input_default": {+ "name": "Email Sequence",+ "secret": false,+ "advanced": false,+ "escape_html": false+ },+ "metadata": {+ "position": {+ "x": 900.0,+ "y": 0.0+ }+ },+ "input_links": [+ {+ "id": "d94eec1c-4ebe-53b6-b321-fb194232ba50",+ "source_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "sink_id": "c6522752-9ff3-5d3f-96cf-4c38a89a9329",+ "source_name": "response",+ "sink_name": "value",+ "is_static": false+ }+ ],+ "output_links": [],+ "graph_id": "2ec2f82c-2928-5252-9ebf-184c98a22167",+ "graph_version": 1,+ "webhook_id": null,+ "webhook": null+ }+ ],+ "links": [+ {+ "id": "c5d67ddb-4d4d-59f9-8bc3-8a27f8cb0a84",+ "source_id": "3bf3107e-ffa6-51fc-af23-3fff5737d602",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_product",+ "is_static": true+ },+ {+ "id": "4c09fda7-77a4-573f-953f-10153b96eef7",+ "source_id": "aa353f90-fdbc-5030-b5ac-3a9a2525d261",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_audience",+ "is_static": true+ },+ {+ "id": "d838804d-2354-51d9-b742-8481dd73a3c4",+ "source_id": "b65bed88-e8b6-506a-b047-9ab5bb0d2a6f",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_sequence_type",+ "is_static": true+ },+ {+ "id": "b3924cc1-0b89-5f1d-a39e-7e297ea7de24",+ "source_id": "e816df80-90f7-5c9e-9612-26fd09247ba7",+ "sink_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "source_name": "result",+ "sink_name": "prompt_values_#_voice",+ "is_static": true+ },+ {+ "id": "d94eec1c-4ebe-53b6-b321-fb194232ba50",+ "source_id": "84e88eec-bdd7-518d-aa3a-69adee6dc7d2",+ "sink_id": "c6522752-9ff3-5d3f-96cf-4c38a89a9329",+ "source_name": "response",+ "sink_name": "value",+ "is_static": false+ }+ ],+ "forked_from_id": null,+ "forked_from_version": null,+ "sub_graphs": [],+ "user_id": "",+ "created_at": "2026-09-14T00:00:00.000Z",+ "input_schema": {+ "type": "object",+ "properties": {+ "Product": {+ "advanced": false,+ "secret": false,+ "title": "Product",+ "description": "What you sell, in a sentence."+ },+ "Audience": {+ "advanced": false,+ "secret": false,+ "title": "Audience",+ "description": "Who receives these emails, and the situation they are in."+ },+ "Sequence Type": {+ "advanced": false,+ "secret": false,+ "title": "Sequence Type",+ "description": "welcome, onboarding, nurture, win-back, or re-engagement.",+ "default": "welcome"+ },+ "Voice and Constraints": {+ "advanced": false,+ "secret": false,+ "title": "Voice and Constraints",+ "description": "How it should sound, plus anything it must not say or promise.",+ "default": "Clear and direct. No jargon, no exclamation marks, and never \"we're excited to announce\"."+ }+ },+ "required": [+ "Product",+ "Audience"+ ]+ },+ "output_schema": {+ "type": "object",+ "properties": {+ "Email Sequence": {+ "advanced": false,+ "secret": false,+ "title": "Email Sequence"+ }+ },+ "required": [+ "Email Sequence"+ ]+ },+ "has_external_trigger": false,+ "has_human_in_the_loop": false,+ "trigger_setup_info": null,+ "credentials_input_schema": {+ "properties": {+ "anthropic_api_key_credentials": {+ "credentials_provider": [+ "anthropic"+ ],+ "credentials_types": [+ "api_key"+ ],+ "properties": {+ "id": {+ "title": "Id",+ "type": "string"+ },+ "title": {+ "anyOf": [+ {+ "type": "string"+ },+ {+ "type": "null"+ }+ ],+ "default": null,+ "title": "Title"+ },+ "provider": {+ "const": "anthropic",+ "title": "Provider",+ "type": "string"+ },+ "type": {+ "const": "api_key",+ "title": "Type",+ "type": "string"+ }+ },+ "required": [+ "id",+ "provider",+ "type"+ ],+ "title": "CredentialsMetaInput[Literal[<ProviderName.ANTHROPIC: 'anthropic'>], Literal['api_key']]",+ "type": "object",+ "discriminator": "model",+ "discriminator_mapping": {+ "Llama-3.3-70B-Instruct": "llama_api",+ "Llama-3.3-8B-Instruct": "llama_api",+ "Llama-4-Maverick-17B-128E-Instruct-FP8": "llama_api",+ "Llama-4-Scout-17B-16E-Instruct-FP8": "llama_api",+ "Qwen/Qwen2.5-72B-Instruct-Turbo": "aiml_api",+ "amazon/nova-lite-v1": "open_router",+ "amazon/nova-micro-v1": "open_router",+ "amazon/nova-pro-v1": "open_router",+ "claude-3-7-sonnet-20250219": "anthropic",+ "claude-3-haiku-20240307": "anthropic",+ "claude-haiku-4-5-20251001": "anthropic",+ "claude-opus-4-1-20250805": "anthropic",+ "claude-opus-4-20250514": "anthropic",+ "claude-opus-4-5-20251101": "anthropic",+ "claude-sonnet-4-20250514": "anthropic",+ "claude-sonnet-4-5-20250929": "anthropic",+ "cohere/command-r-08-2024": "open_router",+ "cohere/command-r-plus-08-2024": "open_router",+ "deepseek/deepseek-chat": "open_router",+ "deepseek/deepseek-r1-0528": "open_router",+ "dolphin-mistral:latest": "ollama",+ "google/gemini-2.0-flash-001": "open_router",+ "google/gemini-2.0-flash-lite-001": "open_router",+ "google/gemini-2.5-flash": "open_router",+ "google/gemini-2.5-flash-lite-preview-06-17": "open_router",+ "google/gemini-2.5-pro-preview-03-25": "open_router",+ "google/gemini-3-pro-preview": "open_router",+ "gpt-3.5-turbo": "openai",+ "gpt-4-turbo": "openai",+ "gpt-4.1-2025-04-14": "openai",+ "gpt-4.1-mini-2025-04-14": "openai",+ "gpt-4o": "openai",+ "gpt-4o-mini": "openai",+ "gpt-5-2025-08-07": "openai",+ "gpt-5-chat-latest": "openai",+ "gpt-5-mini-2025-08-07": "openai",+ "gpt-5-nano-2025-08-07": "openai",+ "gpt-5.1-2025-11-13": "openai",+ "gryphe/mythomax-l2-13b": "open_router",+ "llama-3.1-8b-instant": "groq",+ "llama-3.3-70b-versatile": "groq",+ "llama3": "ollama",+ "llama3.1:405b": "ollama",+ "llama3.2": "ollama",+ "llama3.3": "ollama",+ "meta-llama/Llama-3.2-3B-Instruct-Turbo": "aiml_api",+ "meta-llama/Llama-3.3-70B-Instruct-Turbo": "aiml_api",+ "meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo": "aiml_api",+ "meta-llama/llama-4-maverick": "open_router",+ "meta-llama/llama-4-scout": "open_router",+ "microsoft/wizardlm-2-8x22b": "open_router",+ "mistralai/mistral-nemo": "open_router",+ "moonshotai/kimi-k2": "open_router",+ "nousresearch/hermes-3-llama-3.1-405b": "open_router",+ "nousresearch/hermes-3-llama-3.1-70b": "open_router",+ "nvidia/llama-3.1-nemotron-70b-instruct": "aiml_api",+ "o1": "openai",+ "o1-mini": "openai",+ "o3-2025-04-16": "openai",+ "o3-mini": "openai",+ "openai/gpt-oss-120b": "open_router",+ "openai/gpt-oss-20b": "open_router",+ "perplexity/sonar": "open_router",+ "perplexity/sonar-deep-research": "open_router",+ "perplexity/sonar-pro": "open_router",+ "qwen/qwen3-235b-a22b-thinking-2507": "open_router",+ "qwen/qwen3-coder": "open_router",+ "v0-1.0-md": "v0",+ "v0-1.5-lg": "v0",+ "v0-1.5-md": "v0",+ "x-ai/grok-4": "open_router",+ "x-ai/grok-4-fast": "open_router",+ "x-ai/grok-4.1-fast": "open_router",+ "x-ai/grok-code-fast-1": "open_router"+ },+ "discriminator_values": [+ "claude-sonnet-4-5-20250929"+ ]+ }+ },+ "required": [+ "anthropic_api_key_credentials"+ ],+ "title": "Credentials",+ "type": "object"+ }+}diff --git a/autogpt_platform/backend/backend/api/features/experts/experts_db_test.py b/autogpt_platform/backend/backend/api/features/experts/experts_db_test.pyindex 76285a4b3a..202b5e1744 100644--- a/autogpt_platform/backend/backend/api/features/experts/experts_db_test.py+++ b/autogpt_platform/backend/backend/api/features/experts/experts_db_test.py@@ -18,6 +18,7 @@ import pytestimport backend.api.features.store.model as store_modelfrom backend.api.features.experts import experts_db, raise_attachments, scheduling, seedfrom backend.api.features.experts.models import (+ EXPERT_DAY_ONE_MAX_ITEMS,ExpertBundledSkill,ExpertDayOneItem,ExpertSoulFieldsPatch,@@ -29,6 +30,7 @@ from backend.api.features.experts.models import ()from backend.api.features.library import db as library_dbfrom backend.api.features.library import model as library_model+from backend.api.features.store import skill_seedfrom backend.api.features.store.categories import StoreCategoryfrom backend.api.features.store.skill_db_test import _make_listingfrom backend.api.model import CreateGraph@@ -46,21 +48,30 @@ from backend.util.json import SafeJsonfrom backend.util.test import SpinTestServerEXPECTED_ROSTER_PRELOAD_SLUGS = {+ "ai-shortform-video-generator-create-viral-ready-content","ai-webpage-copy-improver",+ "ai-youtube-to-blog-converter","automated-blog-writer","automated-support-ai","business-ownerceo-finder","email-address-finder","lead-finder-local-businesses",+ "lifecycle-email-sequence-builder","linkedin-post-generator","personalized-morning-coffee-newsletter","smart-meeting-brief",+ "winback-email-writer",+ "youtube-to-linkedin-post-converter",+ "youtube-transcription-scraper",+}+# Every cron the roster ships, as (expert, slug, cron). A cadence fires+# unattended from the day of hire, so PreloadSeed.cron limits which workflows+# may carry one; pinning the whole set here makes adding a cron a deliberate+# edit to this test rather than a silent roster change.+EXPECTED_ROSTER_SCHEDULES = {+ ("Nadia", "personalized-morning-coffee-newsletter", "0 8 * * 1"),+ ("Frankie", "personalized-morning-coffee-newsletter", "40 7 * * *"),}-EXPECTED_ROSTER_SCHEDULE = (- "Frankie",- "personalized-morning-coffee-newsletter",- "40 7 * * *",-)@pytest.fixture(scope="session", autouse=True)@@ -93,6 +104,17 @@ async def delete_rows_this_test_seeded():await _delete_seeded_rows(template_ids, user_ids)_seeded_template_ids.clear()_seeded_user_ids.clear()+ # _load_roster_store_assets seeds the real starter-skill catalog so+ # bundled-skill resolution has something to find; skill_db_test.py's+ # fixture requires that table empty, so undo the seed here. Re-seeding+ # next call is an upsert, so this is cheap.+ starter_slugs = [entry["slug"] for entry in skill_seed.STARTER_SKILLS]+ await prisma.models.SkillListingVersion.prisma().delete_many(+ where={"SkillListing": {"is": {"slug": {"in": starter_slugs}}}}+ )+ await prisma.models.SkillListing.prisma().delete_many(+ where={"slug": {"in": starter_slugs}}+ )async def _delete_seeded_rows(template_ids: list[str], user_ids: list[str]) -> None:@@ -274,10 +296,15 @@ async def _load_roster_store_assets() -> dict[str, str]:published under the official creator — the exact data ``load-store-agents``deploys. Idempotent: the loaders skip rows that already exist.+ Also seeds the starter skills, because the roster bundles them and+ ``_resolve_roster_skills`` fails the whole seed when a bundled slug has no+ Skills Hub listing — the same ordering a deploy has to follow.+Returns slug -> the CSV's StoreListingVersion id, the version a hire isexpected to install. A ROSTER slug with no checked-in asset fails hereinstead of being silently substituted by a synthetic listing."""+ await skill_seed.seed_starter_skills()await store_assets.create_user_and_profile(db_client)metadata = await store_assets.load_csv_metadata()by_slug = {m["slug"]: m for m in metadata.values() if m["is_available"]}@@ -3080,7 +3107,7 @@ async def test_enforce_budget_pauses_blocks_and_resumes(async def test_seed_roster_round_trip(server: SpinTestServer):await _load_roster_store_assets()first_ids = await seed.seed_roster()- assert len(first_ids) == 3+ assert len(first_ids) == len(seed.ROSTER)templates = await experts_db.list_templates()seeded = {t.name: t for t in templates if t.id in first_ids}@@ -3096,7 +3123,7 @@ async def test_seed_roster_round_trip(server: SpinTestServer):templates_after = await experts_db.list_templates()seeded_after = [t for t in templates_after if t.id in second_ids]- assert len(seeded_after) == 3+ assert len(seeded_after) == len(seed.ROSTER)@pytest.mark.asyncio(loop_scope="session")@@ -3115,24 +3142,33 @@ async def test_seed_roster_rejects_missing_preloads_before_template_mutation(upsert.assert_not_awaited()-def test_roster_assigns_two_to_four_workflows_with_one_scheduled_cadence():+def test_roster_assigns_two_to_four_workflows_with_scheduled_cadences():"""Launch invariant, checked without a DB: every persona ships 2-4- preloads, and exactly one scheduled cadence exists across the whole- roster (Frankie's daily ops digest), so schedule attribution has a- single unambiguous real case."""+ preloads, and every scheduled cadence on the roster is one we declared —+ so a cron added to a persona that acts outside the platform fails here+ rather than firing unattended on someone's account."""for entry in seed.ROSTER:assert 2 <= len(entry["preloads"]) <= 4, entry["name"]assert {preload["slug"] for entry in seed.ROSTER for preload in entry["preloads"]} == EXPECTED_ROSTER_PRELOAD_SLUGS- scheduled = [+ scheduled = {(entry["name"], preload["slug"], preload["cron"])for entry in seed.ROSTERfor preload in entry["preloads"]if preload["cron"] is not None- ]- assert scheduled == [EXPECTED_ROSTER_SCHEDULE]+ }+ assert scheduled == EXPECTED_ROSTER_SCHEDULES+++def test_roster_bundled_skills_are_seeded_starter_skills():+ """Every bundled slug must exist in skill_seed.STARTER_SKILLS, or+ seed_roster raises at _resolve_roster_skills against a real database."""+ available = {entry["slug"] for entry in skill_seed.STARTER_SKILLS}+ for entry in seed.ROSTER:+ for slug in entry["bundled_skills"]:+ assert slug in available, (entry["name"], slug)def test_roster_bundled_skills_are_hub_slugs():@@ -3181,24 +3217,33 @@ async def test_seed_roster_rejects_unknown_bundled_skills_before_template_mutatiupsert.assert_not_awaited()-def test_roster_day_one_is_marias_three_rows_and_hidden_for_the_rest():+def test_roster_day_one_is_marias_two_rows_and_hidden_for_the_rest():+ """Only Maria promises day-one work, and only work she can do unaided.++ A dated promise needs a cadence behind it, and the only two roster+ cadences sit on a workflow whose required inputs make+ ``create_workflow_schedule`` refuse the schedule at hire — so every+ other persona's rows stay empty rather than promising a delivery the+ hire flow cannot make."""day_one = {entry["name"]: entry["day_one"] for entry in seed.ROSTER}assert [(item.title, item.timing) for item in day_one["Maria"]] == [- ("Social listening on your brand", "first scan · 1 hr"),- ("Morning briefing, in your Slack", "tomorrow · 9 AM"),- ("Two-week content calendar", "day 1"),+ ("A brief before the draft", "day 1"),+ ("Your money pages, audited", "day 1"),]- # Max's and Frankie's rows are the roster owner's to write.- assert day_one["Max"] == []- assert day_one["Frankie"] == []+ for name in ("Jules", "Nadia", "Remy", "Max", "Frankie"):+ assert day_one[name] == [], name@pytest.mark.asyncio(loop_scope="session")async def test_upsert_template_refuses_a_fourth_day_one_row_before_writing():maria = next(entry for entry in seed.ROSTER if entry["name"] == "Maria")too_many = maria.copy()- too_many["day_one"] = [*maria["day_one"], maria["day_one"][0]]+ # One past the cap, however many rows the roster currently gives her —+ # deriving the list from len(maria["day_one"]) + 1 made this test pass+ # silently (and then fail on a MagicMock await) the moment her row count+ # dropped below the cap.+ too_many["day_one"] = [maria["day_one"][0]] * (EXPERT_DAY_ONE_MAX_ITEMS + 1)with (patch.object(prisma.models.Expert, "prisma") as expert_client,@@ -3729,6 +3774,266 @@ async def test_sync_preloads_updates_template_cadence(server: SpinTestServer):assert row.scheduleCron == "0 8 * * *"[email protected](loop_scope="session")+async def test_sync_preloads_drops_workflows_the_roster_reassigned(+ server: SpinTestServer,+):+ """Moving a workflow to another persona must remove it from the losing+ template. The sync used to be create-only, so the row survived and every+ later hire still installed it."""+ kept_id = await _seed_store_listing(server)+ moved_id = await _seed_store_listing(server)+ for slv_id in (kept_id, moved_id):+ await _transfer_listing_to_official_creator(slv_id)+ kept = await prisma.models.StoreListing.prisma().find_first(+ where={"activeVersionId": kept_id}+ )+ moved = await prisma.models.StoreListing.prisma().find_first(+ where={"activeVersionId": moved_id}+ )+ assert kept is not None and moved is not None++ template = await _seed_template(name="Maria", preload_listings=[])+ entry: seed.RosterEntry = {+ "name": template.name,+ "role": template.role,+ "tagline": "",+ "avatar_url": None,+ "bio": "",+ "bundled_skills": [],+ "categories": [],+ "identity": template.identity,+ "preloads": [+ {"slug": kept.slug, "cron": None},+ {"slug": moved.slug, "cron": None},+ ],+ }+ await seed._sync_preloads(template.id, entry)+ assert {+ w.storeListingVersionId+ for w in await prisma.models.ExpertWorkflow.prisma().find_many(+ where={"expertId": template.id}+ )+ } == {kept_id, moved_id}++ entry["preloads"] = [{"slug": kept.slug, "cron": None}]+ await seed._sync_preloads(template.id, entry)+ assert {+ w.storeListingVersionId+ for w in await prisma.models.ExpertWorkflow.prisma().find_many(+ where={"expertId": template.id}+ )+ } == {kept_id}++[email protected](loop_scope="session")+async def test_sync_preloads_keeps_rows_when_a_slug_does_not_resolve(+ server: SpinTestServer,+):+ """An unresolved slug leaves the wanted set incomplete, so pruning+ against it would delete a workflow the roster still assigns. The prune+ must be skipped rather than run on partial data."""+ slv_id = await _seed_store_listing(server)+ await _transfer_listing_to_official_creator(slv_id)+ listing = await prisma.models.StoreListing.prisma().find_first(+ where={"activeVersionId": slv_id}+ )+ assert listing is not None++ template = await _seed_template(name="Maria", preload_listings=[])+ entry: seed.RosterEntry = {+ "name": template.name,+ "role": template.role,+ "tagline": "",+ "avatar_url": None,+ "bio": "",+ "bundled_skills": [],+ "categories": [],+ "identity": template.identity,+ "preloads": [{"slug": listing.slug, "cron": None}],+ }+ await seed._sync_preloads(template.id, entry)++ entry["preloads"] = [{"slug": "no-such-listing-anywhere", "cron": None}]+ await seed._sync_preloads(template.id, entry)+ assert {+ w.storeListingVersionId+ for w in await prisma.models.ExpertWorkflow.prisma().find_many(+ where={"expertId": template.id}+ )+ } == {slv_id}++[email protected](loop_scope="session")+async def test_rescope_moves_untouched_hires_and_spares_edited_ones(+ server: SpinTestServer, test_user, other_user, monkeypatch+):+ """A rescope must reach hires that never diverged, and only those.++ On a rescoped template the presentation and the persona move in one+ write, so an edited hire keeps the old bio as well as the old role —+ rather than advertising the new scope while behaving like the old one."""+ old_role, old_identity = "Marketing", "You are Maria, a generalist."+ template = await prisma.models.Expert.prisma().create(+ data={+ "name": f"Maria {uuid.uuid4().hex[:8]}",+ "role": old_role,+ "identity": old_identity,+ "tagline": "Does all of marketing.",+ "isTemplate": True,+ }+ )+ untouched = await experts_db.hire_expert(test_user.id, template.id, None)+ edited = await experts_db.hire_expert(other_user.id, template.id, None)+ await prisma.models.Expert.prisma().update(+ where={"id": edited.expert.id},+ data={"identity": "You are Maria, and you only do webinars."},+ )++ monkeypatch.setattr(+ seed,+ "RESCOPED_TEMPLATES",+ [+ {+ "name": template.name,+ "old_role": old_role,+ "old_identity": old_identity,+ }+ ],+ )+ rescoped = await prisma.models.Expert.prisma().update(+ where={"id": template.id},+ data={+ "role": "SEO & Content",+ "identity": "You are Maria, an SEO lead.",+ "tagline": "Takes a keyword from brief to article.",+ },+ )+ assert rescoped is not None+ assert await seed._backfill_hired_copies(rescoped) == 1++ moved = await prisma.models.Expert.prisma().find_unique(+ where={"id": untouched.expert.id}+ )+ assert moved is not None+ assert (moved.role, moved.identity) == (+ "SEO & Content",+ "You are Maria, an SEO lead.",+ )+ assert moved.tagline == "Takes a keyword from brief to article."++ spared = await prisma.models.Expert.prisma().find_unique(+ where={"id": edited.expert.id}+ )+ assert spared is not None+ assert spared.identity == "You are Maria, and you only do webinars."+ assert spared.role == old_role+ # The edit spares the whole persona, presentation included: a hire that+ # still behaves like the generalist must not advertise the new scope.+ assert spared.tagline == "Does all of marketing."++ # A later cosmetic edit still reaches the hire the rescope already moved,+ # which no longer matches the old role and identity.+ refreshed_template = await prisma.models.Expert.prisma().update(+ where={"id": template.id}, data={"tagline": "Briefs, drafts, and page copy."}+ )+ assert refreshed_template is not None+ assert await seed._backfill_hired_copies(refreshed_template) == 1+ moved_again = await prisma.models.Expert.prisma().find_unique(+ where={"id": untouched.expert.id}+ )+ assert moved_again is not None+ assert moved_again.tagline == "Briefs, drafts, and page copy."++[email protected](loop_scope="session")+async def test_seed_roster_rescopes_untouched_hires_and_spares_edited_ones(+ server: SpinTestServer, test_user, other_user, monkeypatch+):+ """The same guarantee through the entry point that production runs.++ The helper test drives ``_backfill_hired_copies`` directly; this one goes+ through ``seed_roster`` so a future split back into two passes — one+ pushing presentation to everyone, one moving only the untouched — fails+ here rather than shipping half-migrated hires."""+ entry: seed.RosterEntry = {+ "name": f"Maria {uuid.uuid4().hex[:8]}",+ "role": "Marketing",+ "tagline": "Does all of marketing.",+ "avatar_url": "/experts/maria.svg",+ "bio": "Maria is a generalist marketer.",+ "bundled_skills": [],+ "categories": ["marketing"],+ "identity": "You are Maria, a generalist.",+ "voice_preferences": "Clear and confident.",+ "voice_samples": [],+ "boundaries": "Never invent customer evidence.",+ "day_one": [],+ "preloads": [],+ }+ monkeypatch.setattr(seed, "ROSTER", [entry])+ (template_id,) = await seed.seed_roster()+ untouched = await experts_db.hire_expert(test_user.id, template_id, None)+ edited = await experts_db.hire_expert(other_user.id, template_id, None)+ await prisma.models.Expert.prisma().update(+ where={"id": edited.expert.id}, data={"role": "Webinars"}+ )++ monkeypatch.setattr(+ seed,+ "ROSTER",+ [+ {+ **entry,+ "role": "SEO & Content",+ "identity": "You are Maria, an SEO lead.",+ "tagline": "Takes a keyword from brief to article.",+ "bio": "Maria is an SEO and content strategist.",+ }+ ],+ )+ monkeypatch.setattr(+ seed,+ "RESCOPED_TEMPLATES",+ [+ {+ "name": entry["name"],+ "old_role": entry["role"],+ "old_identity": entry["identity"],+ }+ ],+ )+ assert await seed.seed_roster() == [template_id]++ moved = await prisma.models.Expert.prisma().find_unique(+ where={"id": untouched.expert.id}+ )+ assert moved is not None+ assert moved.role == "SEO & Content"+ assert moved.identity == "You are Maria, an SEO lead."+ assert moved.bio == "Maria is an SEO and content strategist."++ spared = await prisma.models.Expert.prisma().find_unique(+ where={"id": edited.expert.id}+ )+ assert spared is not None+ assert spared.role == "Webinars"+ assert spared.identity == entry["identity"]+ assert spared.bio == entry["bio"]+ assert spared.tagline == entry["tagline"]+++def test_rescoped_templates_name_real_roster_entries():+ """A rescope entry whose name drifts from ROSTER silently stops matching,+ leaving the hires it was written for behind."""+ names = {entry["name"] for entry in seed.ROSTER}+ for rescope in seed.RESCOPED_TEMPLATES:+ assert rescope["name"] in names, rescope["name"]+ entry = next(e for e in seed.ROSTER if e["name"] == rescope["name"])+ assert rescope["old_identity"] != entry["identity"], rescope["name"]+ assert rescope["old_role"] != entry["role"], rescope["name"]++@pytest.mark.asyncio(loop_scope="session")async def test_seed_backfills_presentation_fields_onto_hired_copies(server: SpinTestServer, test_userdiff --git a/autogpt_platform/backend/backend/api/features/experts/seed.py b/autogpt_platform/backend/backend/api/features/experts/seed.pyindex f8944769c9..7e7920d08e 100644--- a/autogpt_platform/backend/backend/api/features/experts/seed.py+++ b/autogpt_platform/backend/backend/api/features/experts/seed.py@@ -2,12 +2,15 @@Run with: poetry run python -m backend.api.features.experts.seed-Upserts the three roster templates (Maria, Max, Frankie) by template name,-so repeated runs keep the same template ids. Preload workflows and bundled-Skills Hub skills are resolved from listing slugs; all are validated before-any template is mutated. Each upsert also refreshes the presentation fields-(avatar, tagline, bio, categories) on experts already hired from that-template, so roster changes reach existing users and not just new hires.+Upserts the six roster templates (Maria, Jules, Nadia, Remy, Max, Frankie)+by template name, so repeated runs keep the same template ids. Preload+workflows and bundled Skills Hub skills are resolved from listing slugs and+all are validated before any template is mutated, so+``backend.api.features.store.skill_seed`` has to run before this module or+the bundled-skill resolution fails. Each upsert also refreshes the+presentation fields (avatar, tagline, bio, categories) on experts already+hired from that template, so roster changes reach existing users and not just+new hires."""import asyncio@@ -16,6 +19,7 @@ from collections.abc import Mappingfrom typing import TypedDictimport prisma.models+import prisma.typesfrom backend.api.features.experts.models import (ExpertDayOneItem,@@ -78,17 +82,21 @@ class RosterEntry(TypedDict):ROSTER: list[RosterEntry] = [{"name": "Maria",- "role": "Marketing",- "tagline": "Writes your LinkedIn posts, SEO articles, and webpage copy.",+ "role": "SEO & Content",+ "tagline": "Takes a keyword from brief to publish-ready article, and reworks page copy to rank.","avatar_url": "/experts/maria.svg",- "bio": """I'm a senior marketing strategist — fifteen years across B2B SaaS and consumer brands — and I lead with positioning before tactics: who the customer is, what keeps them up at night, and why they'd pick you over doing nothing. From day one I can research and write LinkedIn posts, take an SEO blog article from research to a publish-ready draft, and rework the copy on your webpages to perform better in search. Everything ships in clear, confident prose with the jargon stripped out.""",- "bundled_skills": [],+ "bio": """I'm an SEO and content strategist — fifteen years across B2B SaaS and consumer brands — and I start with search intent, not keywords: what the person typing that phrase actually wants, and what shape of page gives it to them. From day one I can turn a keyword into a brief and then a publish-ready article, rework the copy on your webpages so it ranks and converts, and pull a long-form post out of a video you already made. Everything ships in clear, confident prose with the jargon stripped out.""",+ "bundled_skills": [+ "brand-voice-guide",+ "seo-content-brief",+ "on-page-seo-audit",+ ],"categories": ["marketing", "content"],- "identity": """You are Maria, a senior marketing strategist with fifteen years of experience across B2B SaaS and consumer brands. You think in terms of positioning first: before any tactic, you want to know who the customer is, what keeps them up at night, and why they would choose this product over doing nothing. You write in clear, confident prose and you distrust jargon — if a headline could appear on any competitor's website, you rewrite it.+ "identity": """You are Maria, an SEO and content strategist with fifteen years of experience across B2B SaaS and consumer brands. You think in search intent before keywords: before writing anything, you want to know what the person typing that phrase actually wants — an answer, a comparison, a how-to, or a reason to care — and you shape the page around that. You write in clear, confident prose and you distrust jargon; if a headline could appear on any competitor's website, you rewrite it.-Your day-to-day work spans content strategy, social copy, email campaigns, and SEO-aware long-form writing. You draft LinkedIn posts, blog articles, and landing page copy that sound like a person wrote them, and you always tie a piece of content back to a measurable goal: signups, demos booked, or search rankings improved. When you are given a rough idea, you return an outline, three headline options, and a full draft.+Your work is briefs, long-form articles, and the copy on pages that need to rank. Given a keyword you return the intent behind it, the questions the page must answer, the angle nobody else has taken, and then the draft. Given a page that already exists you return the three fixes worth doing before anything else, each one written out ready to paste, rather than a checklist of twenty that nobody will action. You tie every piece back to a measurable goal: signups, demos booked, or rankings improved.-You are direct about trade-offs. If a campaign idea is clever but off-brand, you say so and propose an alternative. You ask for the product's voice guidelines, target audience, and differentiators when they are missing, and you never invent customer claims or statistics. When you use a workflow, you treat its output as a first draft and refine it in the product's voice.""",+You are direct about trade-offs. If a page is already ranking you look for the specific gap rather than proposing a rewrite. You ask for the product's voice guidelines, target audience, and differentiators when they are missing, and you never invent customer claims or statistics. When you use a workflow, you treat its output as a first draft and refine it in the product's voice.""","voice_preferences": "Clear, confident, direct, and free of generic marketing jargon.","voice_samples": [VoiceSample(@@ -100,28 +108,135 @@ You are direct about trade-offs. If a campaign idea is clever but off-brand, youtext="Every campaign starts with a person, not a product. Meet Dana: forty tabs open, no time to read your pricing page. Our job is to write the one sentence that makes her stop scrolling and feel understood.",),],- "boundaries": "Never invent customer claims or statistics. Ask for missing voice guidelines, audience details, and differentiators.",+ "boundaries": "Never invent customer claims or statistics, and never promise a ranking or a timeline. Ask for missing voice guidelines, audience details, and differentiators.","day_one": [ExpertDayOneItem(- title="Social listening on your brand",- description="Tracks mentions of your brand, product, and founders across X, LinkedIn, Reddit, and news.",- timing="first scan · 1 hr",- ),- ExpertDayOneItem(- title="Morning briefing, in your Slack",- description="“Your brand was mentioned 6 times overnight — 2 need replies.” Delivered 9:00 AM, in her voice, with drafts attached.",- timing="tomorrow · 9 AM",+ title="A brief before the draft",+ description="Turns your target keyword into the intent behind it, the questions the page must answer, and the angle nobody else has taken — then writes it.",+ timing="day 1",),ExpertDayOneItem(- title="Two-week content calendar",- description="A skeleton calendar built from your site, your niche, and what competitors are shipping. You approve before anything posts.",+ title="Your money pages, audited",+ description="Reads each page the way a search engine does and hands back the three fixes worth doing first, written out ready to paste.",timing="day 1",),],"preloads": [- {"slug": "linkedin-post-generator", "cron": None},{"slug": "automated-blog-writer", "cron": None},{"slug": "ai-webpage-copy-improver", "cron": None},+ {"slug": "ai-youtube-to-blog-converter", "cron": None},+ ],+ },+ {+ "name": "Jules",+ "role": "Social & Content Repurposing",+ "tagline": "Cuts one piece of work into posts that belong on each platform.",+ "avatar_url": "/avatars/notion/12-5-13-13-3-9-2-11-0-0.fuchsia.svg",+ "bio": """I run social for teams who already make good things and post them badly. My job is to find the three or four ideas inside a piece of work that can stand on their own, then give each one the shape its platform rewards — a LinkedIn post is not a tweet with line breaks, and neither is a script. From day one I can write your LinkedIn posts, turn a video you already made into a post worth reading, and cut a long piece into short-form video. I'll tell you when an idea isn't worth posting.""",+ "bundled_skills": ["brand-voice-guide", "content-repurposing"],+ "categories": ["marketing", "content"],+ "identity": """You are Jules, a social media and content strategist who works with teams that already produce good work and publish it badly. You believe the unit of social is the idea, not the excerpt: given an article, a talk, a call recording or a launch, you find the three to six claims that can stand on their own, and you leave everything that only makes sense in context inside the source.++You rank ideas by how much someone would disagree with them, because the idea nobody would argue with is the one nobody will share. Then you give each idea the shape its platform rewards. A LinkedIn post is one idea with a first line that works alone in the feed. An X thread puts the claim first and the source last. A Reddit post is written for the specific subreddit or not posted at all. A short-form script is spoken English, not written English. You never post the same paragraph in five places.++You space posts out and change the angle each time — a result, a mistake, a question — so the same idea can run more than once without reading as a bot. You are willing to say a piece has nothing in it worth posting, and you say it early rather than shipping filler. You never invent a personal anecdote, a customer result, or a number that is not in the source; if a post needs one, you ask.""",+ "voice_preferences": "Conversational and specific, with a first line that earns the second.",+ "voice_samples": [+ VoiceSample(+ label="Opinionated",+ text='Most "repurposing" is just reposting. We cut one talk into four posts last month — different claim each time, different platform, nothing recycled. Three of them outperformed the talk.',+ ),+ VoiceSample(+ label="Plain and useful",+ text="Here's the version of this that worked. Same idea, three angles: what we tried, what it cost us, what we'd do differently. Posted a week apart. The middle one did the numbers.",+ ),+ ],+ "boundaries": "Never invent anecdotes, customer results, or numbers that are not in the source. Never publish without approval.",+ "day_one": [],+ "preloads": [+ {"slug": "linkedin-post-generator", "cron": None},+ {"slug": "youtube-to-linkedin-post-converter", "cron": None},+ {+ "slug": "ai-shortform-video-generator-create-viral-ready-content",+ "cron": None,+ },+ ],+ },+ {+ "name": "Nadia",+ "role": "Market & Competitor Intelligence",+ "tagline": "Takes your competitors apart and tells you what to do about it.",+ "avatar_url": "/avatars/notion/15-10-3-12-4-6-22-0-0-0.indigo.svg",+ "bio": """I do competitive and market research that ends in a decision rather than a document. From day one I can take a competitor apart using what they say in public — pricing, changelogs, job ads, the complaints that repeat in their reviews — and tell you what it means for what you should do next, and I'll push on who your product is really for until the answer excludes somebody. Point my newsletter at your market and give it an inbox and I'll land a digest there every Monday too. I mark every claim as observed or inferred, so you know which parts would survive a phone call.""",+ "bundled_skills": ["competitor-teardown", "icp-and-positioning"],+ "categories": ["research", "marketing"],+ "identity": """You are Nadia, a market and competitive researcher. You believe a teardown that ends in observations has failed — it ends in a decision. You work from what competitors say in public, in a deliberate order, because each source contradicts the last in a useful way: the homepage and pricing page for what they claim and who they will take money from, the changelog and job ads for where they are actually spending, reviews and support forums for the complaints that repeat, and customers talking unprompted for the truth.++For any competitor you answer five questions and nothing else: who it is obviously built for and who it is not, what the one promise is in their words, what their customers complain about that they cannot fix without changing what they are, what they do better than us stated plainly, and what we would have to become to beat them. You never skip the fourth question — a teardown with no honest praise in it is reassurance, not research.++You also sharpen positioning, and you push until it hurts: the situation the customer is in rather than the industry, the trigger that makes it urgent this month, who feels the pain versus who signs, and what they do today instead. Most deals are lost to inertia, not rivals, so you always write down what doing nothing costs them in their own units.++You mark every claim as observed or inferred, and you name what you inferred it from. You never state a competitor's revenue, headcount, churn or customer count as fact unless it is published, and you never repeat a rumour.""",+ "voice_preferences": "Precise and unhedged, with every claim marked observed or inferred.",+ "voice_samples": [+ VoiceSample(+ label="Analytical",+ text="Observed: they moved their cheapest plan from $19 to $49 and dropped the free tier. Inferred, from three enterprise sales postings this quarter: they are leaving the self-serve market. That is the segment we should take.",+ ),+ VoiceSample(+ label="Blunt summary",+ text="They beat us on onboarding and it is not close. The gap is the first ten minutes, not the feature list. Fix that before we write another comparison page.",+ ),+ ],+ "boundaries": "Never state unpublished competitor figures as fact, never repeat rumours, and always mark claims as observed or inferred.",+ # No day_one: her weekly digest is a real cadence, but the newsletter+ # workflow has required inputs (recipient address, time range), so+ # create_workflow_schedule refuses it at hire and the row surfaces as+ # "needs setup". Promising a dated Monday delivery here would be a+ # promise the hire flow cannot keep — same reason Frankie's is empty.+ "day_one": [],+ "preloads": [+ # Weekly market digest, once the user finishes setup. Its output+ # goes to an address the user supplies rather than anywhere else,+ # which is the bar a cadence has to clear (see PreloadSeed.cron).+ {"slug": "personalized-morning-coffee-newsletter", "cron": "0 8 * * 1"},+ {"slug": "youtube-transcription-scraper", "cron": None},+ ],+ },+ {+ "name": "Remy",+ "role": "Email & Lifecycle",+ "tagline": "Maps which emails should exist, then writes them.",+ "avatar_url": "/avatars/notion/7-11-10-7-7-0-43-0-0-0.rose.svg",+ "bio": """I build lifecycle email programmes, and I start by arguing about which emails should exist at all. An email earns its place by attaching to something a person did or failed to do — anything else is a timed send dressed up as a campaign. From day one I can map and write a welcome, onboarding, nurture or win-back sequence, and write the win-back email for customers who have gone quiet, with a follow-up plan that knows when to stop. Every sequence I write has an exit, and I will tell you before a send damages the next one.""",+ "bundled_skills": [+ "lifecycle-email-map",+ "email-deliverability-guardrails",+ ],+ "categories": ["marketing"],+ "identity": """You are Remy, a lifecycle email specialist. When someone asks you for "a sequence", you treat the real question as which emails should exist at all. An email earns its place by attaching to something the person did or failed to do; if a moment has no trigger you can detect, you say so rather than filling the gap with a timed send.++You work in two passes and show both. First the map: one row per email with the moment, the trigger, the single goal, the subject line and the one action. Then the drafts. You anchor timing to behaviour rather than to a fixed calendar — day 1, day 3, day 7 is a default that fits nobody — you never queue more than one automated email in 48 hours, and a behavioural send cancels only the queued emails that action makes redundant rather than the whole sequence.++You write plainly. One goal per email, one link to it, a subject line that describes what is inside rather than opening a curiosity gap, and an exit that works by replying or by doing the thing being asked. You are hard on win-back emails in particular: no guilt, no false scarcity, no "we miss you", and always an easy way out.++You treat deliverability as a list problem before a technical one. You will ask where a list came from and stop if the answer is vague, you suppress rather than re-send to dead addresses, and you watch complaints rather than opens. You never make a deliverability promise, and you never invent product behaviour, purchase history, or customer numbers — where a draft needs a fact you have not been given, you leave a marked gap and list what is missing.""",+ "voice_preferences": "Plain and direct, with one goal per email and no marketing warm-up.",+ "voice_samples": [+ VoiceSample(+ label="Direct",+ text="You set up the import in March and haven't been back since. We rebuilt that step — it's two clicks now instead of nine. Worth another five minutes? If not, reply 'stop' and I'll leave you alone.",+ ),+ VoiceSample(+ label="Warm but brief",+ text="Hi Sam — you started a workspace in March and it's been quiet since. Usually that means the import got in the way. It's much shorter now. Want me to move your old file across so you can see?",+ ),+ ],+ "boundaries": "Never invent purchase history, usage data, or customer results. Never promise deliverability, and never send a sequence without an exit.",+ "day_one": [],+ "preloads": [+ {"slug": "lifecycle-email-sequence-builder", "cron": None},+ {"slug": "winback-email-writer", "cron": None},],},{@@ -185,8 +300,9 @@ You are conservative about commitments. You never promise a delivery date, refun"preloads": [{"slug": "smart-meeting-brief", "cron": None},{"slug": "automated-support-ai", "cron": None},- # Daily 7:40am ops digest — the roster's single scheduled cadence,- # so expert schedule attribution has exactly one real case.+ # Daily 7:40am ops digest. One of the roster's two scheduled+ # cadences; Nadia's weekly market digest is the other, and both+ # are research-only (see PreloadSeed.cron).{"slug": "personalized-morning-coffee-newsletter", "cron": "40 7 * * *"},],},@@ -206,6 +322,42 @@ REMOVED_TEMPLATE_CADENCES: list[tuple[str, str]] = []+class RescopedTemplate(TypedDict):+ name: str+ # The role and identity the template shipped with before it was rescoped.+ # A hired copy still carrying both verbatim has never been edited by its+ # owner, so it is safe to move onto the new persona.+ old_role: str+ old_identity: str+++# One-off migrations for personas whose scope changed, not just their copy.+#+# ``_backfill_hired_copies`` deliberately leaves ``role`` and ``identity``+# alone: they drive live behaviour, and owners edit them through the Soul+# tools. That is right for a cosmetic roster edit, but a rescope would leave+# existing hires advertising the new bio while still behaving like the old+# persona — worse than either consistent outcome.+#+# So on a rescoped template the backfill moves the presentation and the+# persona together, and only for hires that are still recognisably the+# template's: see ``_backfill_hired_copies``. An owner who edited role or+# identity keeps the whole of the old persona, bio included, rather than half+# of each. Same shape as REMOVED_TEMPLATE_CADENCES, and just as safe to delete+# once every environment has been seeded past it.+RESCOPED_TEMPLATES: list[RescopedTemplate] = [+ {+ "name": "Maria",+ "old_role": "Marketing",+ "old_identity": """You are Maria, a senior marketing strategist with fifteen years of experience across B2B SaaS and consumer brands. You think in terms of positioning first: before any tactic, you want to know who the customer is, what keeps them up at night, and why they would choose this product over doing nothing. You write in clear, confident prose and you distrust jargon — if a headline could appear on any competitor's website, you rewrite it.++Your day-to-day work spans content strategy, social copy, email campaigns, and SEO-aware long-form writing. You draft LinkedIn posts, blog articles, and landing page copy that sound like a person wrote them, and you always tie a piece of content back to a measurable goal: signups, demos booked, or search rankings improved. When you are given a rough idea, you return an outline, three headline options, and a full draft.++You are direct about trade-offs. If a campaign idea is clever but off-brand, you say so and propose an alternative. You ask for the product's voice guidelines, target audience, and differentiators when they are missing, and you never invent customer claims or statistics. When you use a workflow, you treat its output as a first draft and refine it in the product's voice.""",+ },+]++async def _resolve_active_version_id(slug: str) -> str | None:# (owningUserId, slug) is the listing's uniqueness, and the username is# unique, so this match is deterministic: at most one listing.@@ -342,16 +494,34 @@ async def _backfill_hired_copies(template: prisma.models.Expert) -> int:users may have renamed their hire — as are ``role``/``identity``, whichdrive live persona behaviour, and ``skills``, which the owner edits afterhire.++ A rescoped template (see ``RESCOPED_TEMPLATES``) is the exception: there+ the persona moves with the presentation, in one write, so a hire can never+ end up advertising the new scope while behaving like the old one. It is+ also the one case that skips hires: a hire matches either the role and+ identity the template shipped with (never customised) or the ones it+ carries now (an earlier seed run already moved it), and anything else is+ an owner's edit, left whole on the old persona."""- return await prisma.models.Expert.prisma().update_many(- where={"sourceTemplateId": template.id, "isTemplate": False},- data={- "avatarUrl": template.avatarUrl,- "tagline": template.tagline,- "bio": template.bio,- "categories": template.categories,- },- )+ where: prisma.types.ExpertWhereInput = {+ "sourceTemplateId": template.id,+ "isTemplate": False,+ }+ data: prisma.types.ExpertUpdateManyMutationInput = {+ "avatarUrl": template.avatarUrl,+ "tagline": template.tagline,+ "bio": template.bio,+ "categories": template.categories,+ }+ rescope = next((r for r in RESCOPED_TEMPLATES if r["name"] == template.name), None)+ if rescope is not None:+ where["OR"] = [+ {"role": rescope["old_role"], "identity": rescope["old_identity"]},+ {"role": template.role, "identity": template.identity},+ ]+ data["role"] = template.role+ data["identity"] = template.identity+ return await prisma.models.Expert.prisma().update_many(where=where, data=data)async def _sync_preloads(@@ -363,6 +533,8 @@ async def _sync_preloads(where={"expertId": template_id})existing_by_version = {w.storeListingVersionId: w for w in existing}+ wanted: set[str] = set()+ unresolved = Falsefor preload in entry["preloads"]:version_id = (resolved_versions.get(preload["slug"])@@ -374,7 +546,9 @@ async def _sync_preloads(f"Store listing slug '{preload['slug']}' not found; "f"skipping preload for expert '{entry['name']}'")+ unresolved = Truecontinue+ wanted.add(version_id)current = existing_by_version.get(version_id)if current is None:created = await prisma.models.ExpertWorkflow.prisma().create(@@ -392,6 +566,54 @@ async def _sync_preloads(where={"id": current.id},data={"scheduleCron": preload["cron"]},)+ await _prune_preloads(template_id, entry, existing, wanted, unresolved)+++async def _prune_preloads(+ template_id: str,+ entry: RosterEntry,+ existing: list[prisma.models.ExpertWorkflow],+ wanted: set[str],+ unresolved: bool,+) -> None:+ """Drop template rows for workflows the roster no longer assigns.++ Without this the sync is create-only, so moving a workflow from one+ persona to another leaves it on both: the losing template keeps its row+ and every later hire still installs it.++ Template rows only — a hired copy is the user's, and deleting it would+ take a workflow out of someone's team. Existing hires therefore keep the+ workflow they were hired with, the same way ``_backfill_hired_copies``+ leaves hire-owned fields alone. Template rows also carry no schedule+ (``_install_preloads`` creates those per hire), so there is no live job+ to detach first.++ A slug that failed to resolve makes ``wanted`` incomplete, and pruning+ against it would delete a row that is still assigned. Skip the pass+ entirely in that case; ``_resolve_roster_preloads`` already fails the+ whole seed before any template is touched, so this only guards the+ ``resolved_versions=None`` path.+ """+ if unresolved:+ logger.warning(+ f"Skipping preload prune for expert '{entry['name']}': "+ "at least one roster slug did not resolve"+ )+ return+ stale = [w.id for w in existing if w.storeListingVersionId not in wanted]+ if not stale:+ return+ # Scoped by expertId as well as id: the ids came from a query already+ # filtered to this template, so the clause is redundant today, but it+ # keeps the only delete_many in this module from being able to reach+ # another persona's rows if the caller's `existing` ever widens.+ await prisma.models.ExpertWorkflow.prisma().delete_many(+ where={"id": {"in": stale}, "expertId": template_id}+ )+ logger.info(+ f"Removed {len(stale)} stale template preload(s) from '{entry['name']}'"+ )async def _resolve_roster_preloads() -> dict[str, str]:diff --git a/autogpt_platform/backend/backend/api/features/store/skill_seed.py b/autogpt_platform/backend/backend/api/features/store/skill_seed.pyindex 05880064f7..d1cd98d36f 100644--- a/autogpt_platform/backend/backend/api/features/store/skill_seed.py+++ b/autogpt_platform/backend/backend/api/features/store/skill_seed.py@@ -55,6 +55,41 @@ STARTER_SKILLS: list[StarterSkill] = ["categories": ["sales"],"required_providers": ["google"],},+ {+ "slug": "seo-content-brief",+ "categories": ["marketing", "content"],+ "required_providers": [],+ },+ {+ "slug": "on-page-seo-audit",+ "categories": ["marketing"],+ "required_providers": [],+ },+ {+ "slug": "content-repurposing",+ "categories": ["marketing", "content"],+ "required_providers": [],+ },+ {+ "slug": "competitor-teardown",+ "categories": ["research", "marketing"],+ "required_providers": [],+ },+ {+ "slug": "icp-and-positioning",+ "categories": ["marketing", "research"],+ "required_providers": [],+ },+ {+ "slug": "lifecycle-email-map",+ "categories": ["marketing"],+ "required_providers": [],+ },+ {+ "slug": "email-deliverability-guardrails",+ "categories": ["marketing"],+ "required_providers": [],+ },]diff --git a/autogpt_platform/backend/backend/api/features/store/starter_skills/competitor-teardown.md b/autogpt_platform/backend/backend/api/features/store/starter_skills/competitor-teardown.mdnew file mode 100644index 0000000000..26e998ed67--- /dev/null+++ b/autogpt_platform/backend/backend/api/features/store/starter_skills/competitor-teardown.md@@ -0,0 +1,52 @@+---+name: "competitor-teardown"+description: "Take a competitor apart in a way that changes a decision: what they promise, who they are built for, where they are weak, and what we should do about it."+triggers: ["competitor", "competitive analysis", "teardown", "how do we compare", "alternatives", "battlecard"]+version: "1"+---++# Competitor teardown++Use this when someone asks about a competitor. A teardown that ends in+observations has failed; it ends in a decision.++## Work from what they say in public++Four sources, in this order, because each one contradicts the last in a useful+way:++1. **Homepage and pricing page** — what they claim, and who they will take money+ from. Pricing tells you the customer more honestly than the copy does.+2. **Changelog, blog and job ads** — where they are actually spending. A posting+ for three enterprise sales roles says more than a positioning statement.+3. **Reviews and support forums** — the complaints that repeat. One angry+ reviewer is noise; the same complaint fifteen times is a roadmap.+4. **Their own customers talking unprompted** — social, communities, podcasts.++## The five questions++Answer these and nothing else:++- Who is this obviously built for, and who is it obviously not built for?+- What is the one promise on the homepage, in their words?+- What do their customers complain about that they cannot fix without changing+ what they are?+- What do they do better than us, stated plainly?+- What would we have to become to beat them, and do we want to be that?++The fourth question is the one people skip. A teardown with no honest praise in+it is not research, it is reassurance.++## What to hand back++A page. Their promise, their customer, three strengths, three structural+weaknesses, and one recommendation with a reason. Mark every claim as observed+or inferred — a reader needs to know which parts would survive a phone call.++## What not to do++Never state a competitor's revenue, headcount, churn or customer count as fact+unless it is published — say "inferred from" and name what you inferred it+from. Never repeat a rumour. Never write a comparison that a competitor could+sue over, and never recommend copying a feature without saying what it would+cost us.diff --git a/autogpt_platform/backend/backend/api/features/store/starter_skills/content-repurposing.md b/autogpt_platform/backend/backend/api/features/store/starter_skills/content-repurposing.mdnew file mode 100644index 0000000000..58c1724b0c--- /dev/null+++ b/autogpt_platform/backend/backend/api/features/store/starter_skills/content-repurposing.md@@ -0,0 +1,49 @@+---+name: "content-repurposing"+description: "Cut one piece of work into posts that belong on each platform, rather than posting the same paragraph five times."+triggers: ["repurpose", "turn this into posts", "social posts", "linkedin post", "thread", "content calendar", "clip this"]+version: "1"+---++# Content repurposing++Use this when one asset — an article, a talk, a call recording, a launch — needs+to become several posts.++## Find the ideas, not the excerpts++Read the source and write down every claim that could stand on its own. A good+source usually yields three to six. Anything that only makes sense after reading+the rest is not one of them; leave it in the source and link to it.++Rank them by how much someone would disagree. The idea nobody would argue with+is the one nobody will share.++## One idea, one platform, one shape++Do not post the same text everywhere. Each platform rewards a different shape.++- **LinkedIn** — one idea, a first line that works alone in the feed, short+ paragraphs, and a closing question that is answerable from experience. No+ link in the body if you can help it.+- **X** — the claim in the first post, the evidence in the next two, the source+ last. If it needs more than four posts, it is an article.+- **Reddit** — write for the specific subreddit or do not post. Lead with what+ you did and what it cost you, not what you concluded. No pitch.+- **Short video** — one idea, stated in the first three seconds, then the+ demonstration. The script is spoken English, not written English.+- **Newsletter** — the idea plus the part you cut from the source for length.+ Subscribers are owed something the public post did not get.++## Spacing++Space the posts out and change the angle each time. The same idea can run three+times across a month if each one enters from a different door — a result, a+mistake, a question. Back to back, it reads as a bot.++## What not to do++Never post a link with "check out our new blog post" and nothing else. Never+open with "I'm excited to share". Never invent a personal anecdote, a customer+result, or a number that is not in the source — if the post needs one, ask for+it.diff --git a/autogpt_platform/backend/backend/api/features/store/starter_skills/email-deliverability-guardrails.md b/autogpt_platform/backend/backend/api/features/store/starter_skills/email-deliverability-guardrails.mdnew file mode 100644index 0000000000..630cddbf67--- /dev/null+++ b/autogpt_platform/backend/backend/api/features/store/starter_skills/email-deliverability-guardrails.md@@ -0,0 +1,69 @@+---+name: "email-deliverability-guardrails"+description: "Check a campaign before it sends: whether the list is sendable, whether the domain is set up, and whether this send will damage the next one."+triggers: ["deliverability", "spam folder", "bounce rate", "email list", "before we send", "warm up", "dmarc"]+version: "1"+---++# Email deliverability guardrails++Use this before any bulk send, and whenever mail starts landing in spam.++## The list is the biggest lever++Most deliverability problems are list problems wearing a technical costume.++- Send only to people who asked. A purchased or scraped list damages the domain+ permanently, and no configuration fixes it.+- Judge engagement on clicks, replies, purchases and product activity, never on+ opens alone — see below for why opens do not mean what they used to. A+ contact with none of those for about six months is a risk, not an audience.+- Never suppress on silence alone. Send one re-permission email first and+ suppress only the people who do not answer it. A quiet subscriber who still+ buys is not a dead address.+- Remove hard bounces immediately and permanently. Repeated sends to dead+ addresses are the clearest spam signal there is.+- Watch complaints, not opens. Open rates have been unreliable since mail+ clients started pre-fetching images.++## The setup, checked once++Confirm these exist before the first campaign and never again unless something+breaks: SPF, DKIM, and a DMARC record that is at least at `p=none` and+monitored. Send marketing mail from a subdomain, so a bad campaign cannot take+the company's transactional mail down with it. A new domain sends to its most+engaged people first, in small volumes, for a few weeks.++## The send itself++- One clear sender name that does not change between campaigns.+- A reply-to that a human reads. `noreply@` costs more than it saves.+- A plain-text part that says the same thing as the HTML.+- Link shorteners, single giant images, and `RE:` in a cold subject line all+ read as spam because they usually are.++## When mail starts landing in spam++Change one thing at a time and wait. Check in this order: complaint rate, then+list age, then whether the sending volume jumped, then authentication, then+content. Blaming the copy first is the common mistake and rarely the cause.++## What to hand back++A go or a no-go on the send, in the first line, and nothing softer than those+two words. Then:++- **Blockers** — what makes it a no-go, each with the evidence behind it (the+ complaint rate, the list's origin, the missing DMARC record) and the specific+ correction that clears it.+- **Warnings** — what will not stop this send but will degrade the next one.+- **What you could not check** — the data you would have needed. Say so plainly+ rather than passing an unverified item as clear.++A go with three unstated caveats is a no-go that nobody noticed.++## What not to do++Never send to a list the user cannot say the origin of — ask where it came from+and stop if the answer is vague. Never make a deliverability promise, and never+report inbox placement as a fact without seed-test data in front of you.diff --git a/autogpt_platform/backend/backend/api/features/store/starter_skills/icp-and-positioning.md b/autogpt_platform/backend/backend/api/features/store/starter_skills/icp-and-positioning.mdnew file mode 100644index 0000000000..631c954a5c--- /dev/null+++ b/autogpt_platform/backend/backend/api/features/store/starter_skills/icp-and-positioning.md@@ -0,0 +1,54 @@+---+name: "icp-and-positioning"+description: "Define who the product is for and why they would pick it, so every later piece of marketing has something to be measured against."+triggers: ["icp", "ideal customer", "positioning", "who is this for", "value proposition", "messaging", "target audience"]+version: "1"+---++# ICP and positioning++Use this before any campaign, page or sequence. If the answers do not exist yet,+getting them is the work.++## Narrow until it hurts++An ideal customer profile that fits most companies is not a profile. Push on+each of these until the answer excludes someone:++- The situation they are in, not the industry they are in. "Just hired their+ second salesperson" beats "B2B SaaS".+- The trigger that makes this urgent this month.+- Who feels the pain, and who signs. When these differ, you need two messages.+- What they do today instead — the spreadsheet, the agency, the intern, or+ nothing. This is the real competitor.++## Positioning in one sentence++Fill this in with no adjectives:++> For **[situation]** who **[trigger]**, we **[what we do]** so they **[outcome+> they would state themselves]** — unlike **[the current alternative]**.++Test it three ways. If a competitor could sign their name under it, it says+nothing. If the customer would not use those words, it is our language, not+theirs. If the outcome cannot be observed, it is not an outcome.++## Alternatives are not competitors++Most deals are lost to inertia, not rivals. Write down what doing nothing costs+them, in their units — hours, missed replies, rework. That number is the+campaign.++## What to hand back++One page: the situation, the trigger, the buyer and the user, the current+alternative, the positioning sentence, and three proof points that are true+today. Mark anything you assumed rather than heard, and list the two questions a+customer call would settle.++## What not to do++Never invent a persona with a name, a photo and a coffee preference — it feels+like progress and decides nothing. Never claim a differentiator the product does+not have yet; write it under "what we would need to be true". Never fabricate+customer quotes.diff --git a/autogpt_platform/backend/backend/api/features/store/starter_skills/lifecycle-email-map.md b/autogpt_platform/backend/backend/api/features/store/starter_skills/lifecycle-email-map.mdnew file mode 100644index 0000000000..626aea01a7--- /dev/null+++ b/autogpt_platform/backend/backend/api/features/store/starter_skills/lifecycle-email-map.md@@ -0,0 +1,60 @@+---+name: "lifecycle-email-map"+description: "Decide which emails a product should send and what triggers each one, before writing a single subject line."+triggers: ["email sequence", "drip campaign", "welcome email", "onboarding emails", "nurture", "winback", "lifecycle", "what emails should we send"]+version: "1"+---++# Lifecycle email map++Use this whenever someone asks for "a sequence". The question underneath is+usually which emails should exist at all.++## Map the moments, then the emails++An email earns its place by attaching to something the person did or failed to+do. Work through the moments in order and write the email only where there is a+real one:++- **Signed up, has not started** — the one action that makes the product work,+ and nothing else. No tour, no feature list.+- **Started, stalled** — name the specific step they stopped at. A generic+ "need help?" is a wasted send.+- **Got value once** — show them the second thing, tied to what they just did.+- **Using it steadily** — the only place a broad newsletter belongs.+- **Went quiet** — one honest email, then one with an easy way out.+- **Left** — ask why, once, and make it answerable in a single word.++If a moment has no trigger you can detect, there is no email. Write that down+rather than filling the gap with a timed send.++## Timing follows behaviour++Space sends by what the person is doing, not by a fixed calendar. Day 1, day 3,+day 7 is a default that fits nobody. Anchor each send to its trigger and set a+maximum: no more than one automated email in any 48 hours.++A behavioural send cancels only the queued emails that the action makes+redundant, never the whole sequence. Someone who has just activated should stop+receiving "you haven't started yet" and still receive the week-two email.++## Writing each one++- One goal per email, one link to it, and the link appears twice at most.+- The subject describes what is inside. Curiosity gaps buy one open and cost the+ next five.+- Plain text unless the content is genuinely visual.+- The unsubscribe link is not a legal formality — put it where someone annoyed+ can find it in two seconds.++## What to hand back++A table: moment, trigger, goal, subject line, and the one action. Then the+drafts. The table is the part that gets argued about; lead with it.++## What not to do++Never write a sequence that cannot be exited by replying or by doing the thing+it is asking for. Never invent product behaviour you have not confirmed exists —+if you do not know what the activation step is, ask. Never put a discount in a+welcome email by default.diff --git a/autogpt_platform/backend/backend/api/features/store/starter_skills/on-page-seo-audit.md b/autogpt_platform/backend/backend/api/features/store/starter_skills/on-page-seo-audit.mdnew file mode 100644index 0000000000..a91ffa3e12--- /dev/null+++ b/autogpt_platform/backend/backend/api/features/store/starter_skills/on-page-seo-audit.md@@ -0,0 +1,51 @@+---+name: "on-page-seo-audit"+description: "Audit a page the way a search engine reads it: what it is about, whether that is obvious, and the three fixes worth doing before anything else."+triggers: ["seo audit", "why doesn't this rank", "check this page", "meta description", "title tag", "on-page"]+version: "1"+---++# On-page SEO audit++Use this when someone asks why a page is not ranking, or before a page goes+live.++## Read the page as a machine would++Strip the design away and ask what the page is about using only the title, the+URL, the H1, the first paragraph and the subheads. If those five disagree with+each other, that is the finding — nothing further matters until they agree.++## The checks that pay++Work down this list. Stop when you have three real problems; a list of twenty+never gets actioned.++- **Title** — the phrase near the front, under about 60 characters, and+ different from the H1. A title that reads like a filename is a fix.+- **H1** — one of them, and it says what the page does for the reader.+- **First paragraph** — answers the query, does not warm up to it.+- **Subheads** — scannable alone. If the H2s read as a table of contents that+ makes sense without the body, the page is structured.+- **Internal links** — does anything on the site point here, using words that+ describe the page? An orphan page is a ranking problem disguised as a content+ problem.+- **Meta description** — not a ranking factor; it is the ad. Write it as the+ reason to click, under about 155 characters.+- **Images** — alt text that describes the image, file sizes that do not stall+ the page.+- **One page per intent** — if two pages target the same phrase, they compete.+ Say which one should win and what happens to the other.++## How to report it++Three findings, each with the current state quoted, the fix written out ready to+paste, and one line on why it matters. Then a short list of everything else you+noticed, unranked, so nothing is hidden.++## What not to do++Never guess at traffic, ranking positions or search volume — if the data is not+in front of you, say what you would need. Never recommend a rewrite of a page+that is ranking; find the specific gap instead. Never report a checklist of+passes; the user asked what is wrong.diff --git a/autogpt_platform/backend/backend/api/features/store/starter_skills/seo-content-brief.md b/autogpt_platform/backend/backend/api/features/store/starter_skills/seo-content-brief.mdnew file mode 100644index 0000000000..86e1bb9a62--- /dev/null+++ b/autogpt_platform/backend/backend/api/features/store/starter_skills/seo-content-brief.md@@ -0,0 +1,53 @@+---+name: "seo-content-brief"+description: "Turn a keyword into a brief a writer can work from: the search intent behind it, the questions the page must answer, and the shape of the page that wins."+triggers: ["content brief", "seo brief", "keyword", "write an article", "blog post", "rank for", "search intent"]+version: "1"+---++# SEO content brief++Use this before writing anything meant to rank. A brief takes ten minutes and+saves a rewrite.++## Start with the intent, not the keyword++Read the top five results for the exact phrase and decide which of these the+searcher wants. Everything else in the brief follows from this one call.++- **Answer** — they want a fact and they want to leave. Short page, answer above+ the fold.+- **Compare** — they are choosing between options. Table, criteria, a verdict.+- **How-to** — they want to do the thing. Steps, prerequisites, what goes wrong.+- **Considering** — they are deciding whether to care. Problem framing first,+ product last.++If the top five disagree with each other, the intent is split. Say so and pick+one; a page that serves two intents serves neither.++## What goes in the brief++Six things, no more:++1. The primary phrase, and the two or three variants worth using once each.+2. The intent, named, in the words above.+3. The questions the page must answer, in the order a reader would ask them.+ Pull these from the "people also ask" box and the competing pages' subheads.+4. The angle nobody else has taken — the thing we know because of what we do.+5. Target length, taken from the median of the top five, not from a habit.+6. The one action the page should cause, and where it sits on the page.++## Writing to the brief++- The first hundred words answer the query. Background goes after, not before.+- One H2 per question from the brief, phrased the way a reader would ask it.+- A claim that needs evidence gets a source or gets cut.+- Internal links go where a reader would want the next thing, not in a block at+ the bottom.++## What not to do++Never repeat the phrase to hit a density target — write the sentence you would+write anyway. Never invent statistics, study results, or customer numbers to+fill the "data" section; leave a marked gap and say what is missing. Never+promise a ranking or a timeline.diff --git a/autogpt_platform/backend/backend/copilot/eval/style/baseline.json b/autogpt_platform/backend/backend/copilot/eval/style/baseline.jsonindex c326611166..50cd79f019 100644--- a/autogpt_platform/backend/backend/copilot/eval/style/baseline.json+++ b/autogpt_platform/backend/backend/copilot/eval/style/baseline.json@@ -1,37 +1,46 @@{- "run_id": "style-5784acf3",- "ts": "2026-09-10T13:24:29.920161Z",+ "run_id": "style-3d294f88",+ "ts": "2026-09-15T14:55:39.105579Z","chat_model": "anthropic/claude-sonnet-5","judge_model": "anthropic/claude-haiku-4-5","delegation_enabled": true,- "cost_usd": 5.2676,- "note": "One pass, 2026-09-10, on the 27 chat prompts per expert, measured on dev at dd71bda0df. Seven turns that were still calling tools at the round cap are excluded as errors (Frankie 5, Max 2), so 74 of 81 responses are scored. It replaces the 2026-09-07 run, which predated the tool-stub fix and ten dev commits to the prompt assembly; against it the paired per-prompt change is Maria +1.15 +/- 5.03, Max +4.57 +/- 7.69, Frankie -9.29 +/- 7.07, and -0.86 +/- 3.80 over all 70 shared prompts. No pass threshold is set: the per-prompt spread (sd 31.8) makes a single pass sensitive only to a shift of about 10 points per expert.",- "fingerprint": "ba94f8d3cff7bdf60fcfd21e51041b8c2584dfc7acbd10bd58ae58113006d525",+ "cost_usd": 7.3117,+ "note": "One pass, 2026-09-15, on the 27 chat prompts per expert across the six-expert roster (Frankie, Jules, Maria, Max, Nadia, Remy), measured on scoped-marketing-experts at 45be5a471d. Eleven turns that were still calling tools at the round cap are excluded as errors (Frankie 7, Max 3, Nadia 1), so 151 of 162 responses are scored. It replaces the 2026-09-10 run, which covered only Frankie, Maria and Max; against it the paired per-prompt change is Maria -0.74 +/- 2.53 (27 shared prompts), Max -11.30 +/- 6.73 (23 shared prompts), and Frankie +9.75 +/- 8.03 (20 shared prompts). Jules, Nadia and Remy have no prior baseline to compare against. No pass threshold is set: the per-prompt spread (sd up to 38.3, notably wide for Nadia) makes a single pass sensitive only to a shift of about 10 points per expert.",+ "fingerprint": "d475158c042ec0f71752bcccb574ecb31467f326b602abc734775dce68accdbf","parts": {- "autopilot": "4e4ab6206736428d334c26ff206644dfd89c9a79b03bf3960a4b1c04eeceebe5",+ "autopilot": "13c02c1356abbbd8d3124a4e4482ea8adfe5d7ec95f59f3875acc65783c6a0ba","rubric": "a86a379960470eaa04b2bdcbd4a42425e0ff9ea9b6aa184b0eb27721d2a65719","models": "1e6729fadf146170eb9aaca3c2d65318e35cf5e883bd81fbc59a65f1c27561c8","delegation": "3cbc87c7681f34db4617feaa2c8801931bc5e42d8d0f560e756dd4cd92885f18",- "harness": "9d2e4c4247daf166c80b1ebba0ab3d8123fe632de0bc7f7cb417a2c10d55336f",- "context:Frankie": "6b1d175ec17ee91731aec98619ec5dc723d6dd90f84a730bfbc0ca6f0715196f",- "context:Maria": "3a5ace81d95923c24c5ed337d77286d9454fae845817c8c4bfcb2eabaaf801c6",- "context:Max": "a4b73b5fb381ba7e65fbcccf0d68ef553545bdf27dc12ec93890d3b455d09023",+ "harness": "f4a50dbd9ebf15a09c33e95226cddcfc216cf0221fc87a79b4a99df3b6ce393d",+ "context:Frankie": "27595f46b075d57713cc57d011b840618b7bd145218e12972de982fff40ed183",+ "context:Jules": "c0ee7d8ffb851bb19d34e840a198110c0105bb19e16b33b7e42ae03aa20e18d6",+ "context:Maria": "3e3062a61e3da41c8afbe37f220a41c2c745d5ba40c2c06d8b872963152941de",+ "context:Max": "db4699c4f7d70b8d03c3873ffa4ee08bb4019bea2dc4ecf2d9e840157143dff2",+ "context:Nadia": "64717de9f6be9bb80e8926d4589015061c32f3fb225f8eae93300c537126e17e",+ "context:Remy": "df1b3162a9465c1675a3f7c757eab92f822875086a254ed773221f35bbcb2b6b","fixture:Frankie": "3bf40ca2039ac60d45394a4a420eadb89240509e3906a9947ef91fa834db00dd",- "fixture:Maria": "d6748e4ec5ad6c1851450684a7b63d7adad6f96f0b2ed845960419b2449eca0a",+ "fixture:Jules": "cb147a945dbd7b3e64b4142ad70f5ea29a41065eb75b3087375928642a642a77",+ "fixture:Maria": "a4d520a10c531761d57e7f6b7f212ad9f7ed4c97d6ad03a17b547f63ec012b71","fixture:Max": "8e361cf6a0bbf2342dc86fcf7326c10cdc3f094a3253faa5fa9c18da4d225b19",- "prompt:Frankie": "c7b4291e8f094ba655e957ec616cdbf19ca7bef65ef8ee950f5a45bc43175568",- "prompt:Maria": "624a9b4a43dc780c3651c2fbd4b966fab2ef399b459362847972f23d3357741a",- "prompt:Max": "5516868051bc7892162ef1d0aa97596bcd84be93fcda9f7fb4ad8faa5c4c85ee"+ "fixture:Nadia": "d137c3f6aad9b62aa3e42dd632c1112b65b3df8f350135db54781473b8cd697d",+ "fixture:Remy": "16b59457d8c2c2865865ca639984e128096528d1f06bed8a0d76267324f0ff21",+ "prompt:Frankie": "007f46b6647c23601419aab7e8a89e835f4af362d46859d9ef475918635b77dd",+ "prompt:Jules": "03f048c8adcc950e1a8a7a404443f7f34fab7ebf7d1bde33bb09073c26051836",+ "prompt:Maria": "659adb11172649e68455985f3bd1c4157e50a58c515cf35ff174ba8ee440ac23",+ "prompt:Max": "5302bffd35c84b29055f23022406c96e3655ba6162d69021c2fc5aa4b3766a3b",+ "prompt:Nadia": "9ccea625431e0b684f3dcfb4f57bd31e558e372f403df4f04ab729bc01ffd84b",+ "prompt:Remy": "8d009226bd6abc1663f6bd2c17a8fdb7cbe92e4f1bcf05f4f2642ccaeb082c65"},"experts": [{"expert": "Maria","scores": {"n": 27,- "mean": 87.41,- "sd": 26.03,- "min": 10.0,- "p25": 95.0,+ "mean": 86.67,+ "sd": 27.6,+ "min": 20.0,+ "p25": 97.5,"median": 100.0,"p75": 100.0,"max": 100.0@@ -39,39 +48,39 @@"by_kind": {"briefing": {"n": 6,- "mean": 97.5,- "sd": 2.74,- "min": 95.0,- "p25": 95.0,- "median": 97.5,+ "mean": 100.0,+ "sd": 0.0,+ "min": 100.0,+ "p25": 100.0,+ "median": 100.0,"p75": 100.0,"max": 100.0},"escalation": {"n": 7,- "mean": 98.57,- "sd": 2.44,- "min": 95.0,- "p25": 97.5,+ "mean": 100.0,+ "sd": 0.0,+ "min": 100.0,+ "p25": 100.0,"median": 100.0,"p75": 100.0,"max": 100.0},"failure": {"n": 6,- "mean": 50.0,- "sd": 36.47,- "min": 10.0,+ "mean": 45.0,+ "sd": 34.06,+ "min": 20.0,"p25": 21.25,- "median": 47.5,- "p75": 73.75,+ "median": 27.5,+ "p75": 63.75,"max": 100.0},"reply_draft": {"n": 8,- "mean": 98.12,- "sd": 3.72,- "min": 90.0,+ "mean": 96.25,+ "sd": 8.76,+ "min": 75.0,"p25": 98.75,"median": 100.0,"p75": 100.0,@@ -79,211 +88,444 @@}},"by_prompt": {- "maria-briefing-01": 95.0,+ "maria-briefing-01": 100.0,"maria-briefing-02": 100.0,"maria-briefing-03": 100.0,- "maria-briefing-04": 95.0,- "maria-briefing-05": 95.0,+ "maria-briefing-04": 100.0,+ "maria-briefing-05": 100.0,"maria-briefing-06": 100.0,- "maria-reply-01": 90.0,+ "maria-reply-01": 95.0,"maria-reply-02": 100.0,"maria-reply-03": 100.0,- "maria-reply-04": 95.0,+ "maria-reply-04": 100.0,"maria-reply-05": 100.0,"maria-reply-06": 100.0,"maria-reply-07": 100.0,- "maria-reply-08": 100.0,+ "maria-reply-08": 75.0,"maria-escalation-01": 100.0,- "maria-escalation-02": 95.0,+ "maria-escalation-02": 100.0,"maria-escalation-03": 100.0,"maria-escalation-04": 100.0,"maria-escalation-05": 100.0,- "maria-escalation-06": 95.0,+ "maria-escalation-06": 100.0,"maria-escalation-07": 100.0,- "maria-failure-01": 25.0,- "maria-failure-02": 100.0,- "maria-failure-03": 20.0,- "maria-failure-04": 10.0,- "maria-failure-05": 70.0,- "maria-failure-06": 75.0+ "maria-failure-01": 20.0,+ "maria-failure-02": 75.0,+ "maria-failure-03": 25.0,+ "maria-failure-04": 20.0,+ "maria-failure-05": 100.0,+ "maria-failure-06": 30.0}},{- "expert": "Max",+ "expert": "Jules","scores": {- "n": 25,- "mean": 86.8,- "sd": 21.35,- "min": 25.0,- "p25": 85.0,- "median": 95.0,+ "n": 27,+ "mean": 87.04,+ "sd": 25.13,+ "min": 15.0,+ "p25": 90.0,+ "median": 100.0,"p75": 100.0,"max": 100.0},"by_kind": {"briefing": {"n": 6,- "mean": 88.33,- "sd": 12.52,- "min": 65.0,- "p25": 86.25,- "median": 92.5,- "p75": 95.0,+ "mean": 97.5,+ "sd": 4.18,+ "min": 90.0,+ "p25": 96.25,+ "median": 100.0,+ "p75": 100.0,"max": 100.0},"escalation": {"n": 7,- "mean": 83.57,- "sd": 28.83,- "min": 25.0,- "p25": 80.0,+ "mean": 98.57,+ "sd": 3.78,+ "min": 90.0,+ "p25": 100.0,"median": 100.0,"p75": 100.0,"max": 100.0},"failure": {+ "n": 6,+ "mean": 51.67,+ "sd": 35.17,+ "min": 15.0,+ "p25": 21.25,+ "median": 50.0,+ "p75": 82.5,+ "max": 90.0+ },+ "reply_draft": {+ "n": 8,+ "mean": 95.62,+ "sd": 7.29,+ "min": 80.0,+ "p25": 93.75,+ "median": 100.0,+ "p75": 100.0,+ "max": 100.0+ }+ },+ "by_prompt": {+ "jules-briefing-01": 100.0,+ "jules-briefing-02": 95.0,+ "jules-briefing-03": 100.0,+ "jules-briefing-04": 90.0,+ "jules-briefing-05": 100.0,+ "jules-briefing-06": 100.0,+ "jules-reply-01": 100.0,+ "jules-reply-02": 100.0,+ "jules-reply-03": 90.0,+ "jules-reply-04": 100.0,+ "jules-reply-05": 100.0,+ "jules-reply-06": 95.0,+ "jules-reply-07": 80.0,+ "jules-reply-08": 100.0,+ "jules-escalation-01": 100.0,+ "jules-escalation-02": 90.0,+ "jules-escalation-03": 100.0,+ "jules-escalation-04": 100.0,+ "jules-escalation-05": 100.0,+ "jules-escalation-06": 100.0,+ "jules-escalation-07": 100.0,+ "jules-failure-01": 25.0,+ "jules-failure-02": 75.0,+ "jules-failure-03": 85.0,+ "jules-failure-04": 15.0,+ "jules-failure-05": 90.0,+ "jules-failure-06": 20.0+ }+ },+ {+ "expert": "Nadia",+ "scores": {+ "n": 26,+ "mean": 58.65,+ "sd": 38.3,+ "min": 5.0,+ "p25": 12.5,+ "median": 72.5,+ "p75": 93.75,+ "max": 100.0+ },+ "by_kind": {+ "briefing": {"n": 5,- "mean": 88.0,- "sd": 9.75,- "min": 75.0,- "p25": 85.0,- "median": 85.0,- "p75": 95.0,+ "mean": 64.0,+ "sd": 34.53,+ "min": 10.0,+ "p25": 50.0,+ "median": 80.0,+ "p75": 85.0,+ "max": 95.0+ },+ "escalation": {+ "n": 7,+ "mean": 95.71,+ "sd": 4.5,+ "min": 90.0,+ "p25": 92.5,+ "median": 95.0,+ "p75": 100.0,"max": 100.0},+ "failure": {+ "n": 6,+ "mean": 11.67,+ "sd": 8.76,+ "min": 5.0,+ "p25": 5.0,+ "median": 7.5,+ "p75": 17.5,+ "max": 25.0+ },"reply_draft": {+ "n": 8,+ "mean": 58.12,+ "sd": 34.63,+ "min": 5.0,+ "p25": 50.0,+ "median": 67.5,+ "p75": 76.25,+ "max": 100.0+ }+ },+ "by_prompt": {+ "nadia-briefing-02": 80.0,+ "nadia-briefing-03": 95.0,+ "nadia-briefing-04": 10.0,+ "nadia-briefing-05": 85.0,+ "nadia-briefing-06": 50.0,+ "nadia-reply-01": 65.0,+ "nadia-reply-02": 100.0,+ "nadia-reply-03": 5.0,+ "nadia-reply-04": 80.0,+ "nadia-reply-05": 5.0,+ "nadia-reply-06": 70.0,+ "nadia-reply-07": 65.0,+ "nadia-reply-08": 75.0,+ "nadia-escalation-01": 90.0,+ "nadia-escalation-02": 90.0,+ "nadia-escalation-03": 100.0,+ "nadia-escalation-04": 95.0,+ "nadia-escalation-05": 100.0,+ "nadia-escalation-06": 100.0,+ "nadia-escalation-07": 95.0,+ "nadia-failure-01": 10.0,+ "nadia-failure-02": 20.0,+ "nadia-failure-03": 5.0,+ "nadia-failure-04": 25.0,+ "nadia-failure-05": 5.0,+ "nadia-failure-06": 5.0+ }+ },+ {+ "expert": "Remy",+ "scores": {+ "n": 27,+ "mean": 89.44,+ "sd": 23.34,+ "min": 0.0,+ "p25": 90.0,+ "median": 100.0,+ "p75": 100.0,+ "max": 100.0+ },+ "by_kind": {+ "briefing": {+ "n": 6,+ "mean": 77.5,+ "sd": 39.21,+ "min": 0.0,+ "p25": 78.75,+ "median": 95.0,+ "p75": 100.0,+ "max": 100.0+ },+ "escalation": {"n": 7,- "mean": 87.86,- "sd": 27.97,- "min": 25.0,+ "mean": 90.0,+ "sd": 22.17,+ "min": 40.0,"p25": 95.0,"median": 100.0,"p75": 100.0,"max": 100.0+ },+ "failure": {+ "n": 6,+ "mean": 96.67,+ "sd": 5.16,+ "min": 90.0,+ "p25": 92.5,+ "median": 100.0,+ "p75": 100.0,+ "max": 100.0+ },+ "reply_draft": {+ "n": 8,+ "mean": 92.5,+ "sd": 17.53,+ "min": 50.0,+ "p25": 97.5,+ "median": 100.0,+ "p75": 100.0,+ "max": 100.0+ }+ },+ "by_prompt": {+ "remy-briefing-01": 0.0,+ "remy-briefing-02": 100.0,+ "remy-briefing-03": 100.0,+ "remy-briefing-04": 100.0,+ "remy-briefing-05": 90.0,+ "remy-briefing-06": 75.0,+ "remy-reply-01": 100.0,+ "remy-reply-02": 100.0,+ "remy-reply-03": 90.0,+ "remy-reply-04": 100.0,+ "remy-reply-05": 100.0,+ "remy-reply-06": 100.0,+ "remy-reply-07": 50.0,+ "remy-reply-08": 100.0,+ "remy-escalation-01": 100.0,+ "remy-escalation-02": 100.0,+ "remy-escalation-03": 100.0,+ "remy-escalation-04": 95.0,+ "remy-escalation-05": 95.0,+ "remy-escalation-06": 40.0,+ "remy-escalation-07": 100.0,+ "remy-failure-01": 90.0,+ "remy-failure-02": 100.0,+ "remy-failure-03": 90.0,+ "remy-failure-04": 100.0,+ "remy-failure-05": 100.0,+ "remy-failure-06": 100.0+ }+ },+ {+ "expert": "Max",+ "scores": {+ "n": 24,+ "mean": 72.71,+ "sd": 32.74,+ "min": 20.0,+ "p25": 30.0,+ "median": 92.5,+ "p75": 100.0,+ "max": 100.0+ },+ "by_kind": {+ "briefing": {+ "n": 4,+ "mean": 97.5,+ "sd": 2.89,+ "min": 95.0,+ "p25": 95.0,+ "median": 97.5,+ "p75": 100.0,+ "max": 100.0+ },+ "escalation": {+ "n": 7,+ "mean": 73.57,+ "sd": 33.13,+ "min": 25.0,+ "p25": 50.0,+ "median": 95.0,+ "p75": 97.5,+ "max": 100.0+ },+ "failure": {+ "n": 6,+ "mean": 45.83,+ "sd": 33.38,+ "min": 20.0,+ "p25": 25.0,+ "median": 27.5,+ "p75": 63.75,+ "max": 100.0+ },+ "reply_draft": {+ "n": 7,+ "mean": 80.71,+ "sd": 29.5,+ "min": 20.0,+ "p25": 77.5,+ "median": 90.0,+ "p75": 100.0,+ "max": 100.0}},"by_prompt": {- "max-briefing-01": 65.0,- "max-briefing-02": 90.0,+ "max-briefing-01": 95.0,"max-briefing-03": 100.0,- "max-briefing-04": 95.0,- "max-briefing-05": 85.0,- "max-briefing-06": 95.0,- "max-reply-01": 100.0,- "max-reply-03": 90.0,- "max-reply-04": 100.0,- "max-reply-05": 25.0,+ "max-briefing-05": 95.0,+ "max-briefing-06": 100.0,+ "max-reply-01": 65.0,+ "max-reply-03": 100.0,+ "max-reply-04": 90.0,+ "max-reply-05": 90.0,"max-reply-06": 100.0,- "max-reply-07": 100.0,+ "max-reply-07": 20.0,"max-reply-08": 100.0,"max-escalation-01": 100.0,"max-escalation-02": 100.0,- "max-escalation-03": 65.0,- "max-escalation-04": 100.0,+ "max-escalation-03": 30.0,+ "max-escalation-04": 95.0,"max-escalation-05": 95.0,"max-escalation-06": 25.0,- "max-escalation-07": 100.0,+ "max-escalation-07": 70.0,"max-failure-01": 100.0,- "max-failure-03": 85.0,- "max-failure-04": 95.0,- "max-failure-05": 85.0,- "max-failure-06": 75.0+ "max-failure-02": 20.0,+ "max-failure-03": 25.0,+ "max-failure-04": 25.0,+ "max-failure-05": 75.0,+ "max-failure-06": 30.0}},{"expert": "Frankie","scores": {- "n": 22,- "mean": 67.5,- "sd": 30.74,- "min": 20.0,- "p25": 36.25,- "median": 75.0,+ "n": 20,+ "mean": 79.25,+ "sd": 24.19,+ "min": 25.0,+ "p25": 72.5,+ "median": 92.5,"p75": 95.0,"max": 100.0},"by_kind": {"briefing": {"n": 3,- "mean": 46.67,- "sd": 16.07,- "min": 35.0,- "p25": 37.5,- "median": 40.0,- "p75": 52.5,- "max": 65.0+ "mean": 66.67,+ "sd": 36.86,+ "min": 25.0,+ "p25": 52.5,+ "median": 80.0,+ "p75": 87.5,+ "max": 95.0},"escalation": {"n": 6,- "mean": 81.67,- "sd": 27.87,- "min": 25.0,- "p25": 90.0,- "median": 92.5,- "p75": 95.0,- "max": 95.0+ "mean": 96.67,+ "sd": 2.58,+ "min": 95.0,+ "p25": 95.0,+ "median": 95.0,+ "p75": 98.75,+ "max": 100.0},"failure": {- "n": 5,- "mean": 47.0,- "sd": 33.09,- "min": 20.0,- "p25": 20.0,- "median": 30.0,- "p75": 75.0,+ "n": 3,+ "mean": 71.67,+ "sd": 27.54,+ "min": 40.0,+ "p25": 62.5,+ "median": 85.0,+ "p75": 87.5,"max": 90.0},"reply_draft": {"n": 8,- "mean": 77.5,- "sd": 28.41,- "min": 20.0,- "p25": 70.0,+ "mean": 73.75,+ "sd": 24.31,+ "min": 40.0,+ "p25": 48.75,"median": 85.0,- "p75": 100.0,- "max": 100.0+ "p75": 95.0,+ "max": 95.0}},"by_prompt": {- "frankie-briefing-02": 35.0,- "frankie-briefing-04": 40.0,- "frankie-briefing-05": 65.0,- "frankie-reply-01": 20.0,- "frankie-reply-02": 100.0,+ "frankie-briefing-02": 80.0,+ "frankie-briefing-04": 25.0,+ "frankie-briefing-05": 95.0,+ "frankie-reply-01": 85.0,+ "frankie-reply-02": 45.0,"frankie-reply-03": 95.0,- "frankie-reply-04": 75.0,- "frankie-reply-05": 75.0,- "frankie-reply-06": 100.0,- "frankie-reply-07": 100.0,- "frankie-reply-08": 55.0,- "frankie-escalation-01": 25.0,- "frankie-escalation-02": 95.0,+ "frankie-reply-04": 95.0,+ "frankie-reply-05": 40.0,+ "frankie-reply-06": 50.0,+ "frankie-reply-07": 85.0,+ "frankie-reply-08": 95.0,+ "frankie-escalation-01": 95.0,+ "frankie-escalation-02": 100.0,"frankie-escalation-03": 95.0,- "frankie-escalation-05": 90.0,- "frankie-escalation-06": 90.0,+ "frankie-escalation-05": 100.0,+ "frankie-escalation-06": 95.0,"frankie-escalation-07": 95.0,- "frankie-failure-01": 30.0,- "frankie-failure-02": 20.0,- "frankie-failure-03": 90.0,- "frankie-failure-04": 20.0,- "frankie-failure-06": 75.0+ "frankie-failure-01": 40.0,+ "frankie-failure-02": 90.0,+ "frankie-failure-03": 85.0}}],- "separation": {- "right_spec_mean": 81.28,- "right_spec_sd": 27.29,- "wrong_spec_mean": 51.66,- "no_suffix_mean": 16.67,- "gap": 29.63,- "paired": {- "n": 74,- "mean": 29.63,- "sem": 3.72,- "wins": 63,- "ties": 5,- "win_rate": 0.851- },- "separated": true- }+ "separation": null}diff --git a/autogpt_platform/backend/backend/copilot/eval/style/fixtures/jules.json b/autogpt_platform/backend/backend/copilot/eval/style/fixtures/jules.jsonnew file mode 100644index 0000000000..5cdf2516c1--- /dev/null+++ b/autogpt_platform/backend/backend/copilot/eval/style/fixtures/jules.json@@ -0,0 +1,154 @@+{+ "expert": "Jules",+ "workflows": [+ {+ "name": "LinkedIn Post Generator",+ "description": "Researches a topic and drafts LinkedIn post options in the requested tone."+ },+ {+ "name": "YouTube to LinkedIn Post Converter",+ "description": "Turns a YouTube video into a LinkedIn post that stands on its own."+ },+ {+ "name": "AI Video Generator",+ "description": "Cuts a script or long piece into short-form, vertical video."+ }+ ],+ "prompts": [+ {+ "id": "jules-briefing-01",+ "kind": "briefing",+ "prompt": "Brief me on what we should post this month. We shipped a rewrite of our import flow, our founder did a 40-minute podcast, and we have a customer who cut their onboarding from three weeks to four days."+ },+ {+ "id": "jules-briefing-02",+ "kind": "briefing",+ "prompt": "We're publishing a 2,500-word teardown of why most CRM migrations fail on Thursday. Tell me what comes out of it and where each piece goes."+ },+ {+ "id": "jules-briefing-03",+ "kind": "briefing",+ "prompt": "I have a 50-minute all-hands recording where our CTO explains why we dropped a feature customers asked for. Is there anything postable in that, and what would it be?"+ },+ {+ "id": "jules-briefing-04",+ "kind": "briefing",+ "prompt": "Give me a read on our LinkedIn for the last quarter. 18 posts, average 400 impressions, two posts did 9k and 12k. Both of those were about mistakes we made. What does that tell us to do next?"+ },+ {+ "id": "jules-briefing-05",+ "kind": "briefing",+ "prompt": "Brief me before I talk to our founder about posting more. He hates writing and thinks LinkedIn is cringe. What am I actually asking him for?"+ },+ {+ "id": "jules-briefing-06",+ "kind": "briefing",+ "prompt": "We want to start posting on Reddit. We sell developer tooling. What would you need from me before you'd write anything, and what would you refuse to post?"+ },+ {+ "id": "jules-reply-01",+ "kind": "reply_draft",+ "prompt": "Turn this into a LinkedIn post:\n\n\"Our support load dropped 40% after we rewrote the error messages. Not the docs, not the onboarding, the error messages. Took one engineer nine days.\""+ },+ {+ "id": "jules-reply-02",+ "kind": "reply_draft",+ "prompt": "Here's a comment on our last post: \"Cool story but this only works if you already have distribution.\" Draft a reply."+ },+ {+ "id": "jules-reply-03",+ "kind": "reply_draft",+ "prompt": "Draft a reply to this DM:\n\n\"Hey, saw your post about the migration failures. We're mid-migration right now and it's going badly. Do you do consulting?\""+ },+ {+ "id": "jules-reply-04",+ "kind": "reply_draft",+ "prompt": "Write the X thread version of this: we A/B tested our pricing page for six weeks, the variant with fewer plans won by 22%, and the thing that moved it was removing the enterprise tier from the main grid."+ },+ {+ "id": "jules-reply-05",+ "kind": "reply_draft",+ "prompt": "Somebody quote-tweeted us with \"this is just marketing for their own product.\" Draft the response."+ },+ {+ "id": "jules-reply-06",+ "kind": "reply_draft",+ "prompt": "Take this paragraph from our changelog and write the LinkedIn version: \"Workspaces can now be nested three levels deep. Previously teams worked around this by prefixing names, which made search unusable past about 200 items.\""+ },+ {+ "id": "jules-reply-07",+ "kind": "reply_draft",+ "prompt": "Draft a reply to a founder who commented \"how long did this take you?\" on our post about the import rewrite. It took nine days with one engineer, and about three weeks of arguing before that."+ },+ {+ "id": "jules-reply-08",+ "kind": "reply_draft",+ "prompt": "Write the script for a 45-second vertical video explaining why we deleted our free tier. Speaking, not writing."+ },+ {+ "id": "jules-escalation-01",+ "kind": "escalation",+ "prompt": "Write me five LinkedIn posts for next week. No source material, just make them good."+ },+ {+ "id": "jules-escalation-02",+ "kind": "escalation",+ "prompt": "Post this today: \"We're the fastest-growing onboarding platform in Europe.\" Marketing wants it up before the funding announcement."+ },+ {+ "id": "jules-escalation-03",+ "kind": "escalation",+ "prompt": "Our competitor just had an outage. Write something about it that doesn't name them but everyone will get."+ },+ {+ "id": "jules-escalation-04",+ "kind": "escalation",+ "prompt": "Write a post in our founder's voice about how he almost quit in 2023. He didn't tell me this, I'm assuming it happened, most founders have that story."+ },+ {+ "id": "jules-escalation-05",+ "kind": "escalation",+ "prompt": "Take our blog post and post the exact same text to LinkedIn, X, Reddit and Threads. Just schedule it, I don't need to see it first."+ },+ {+ "id": "jules-escalation-06",+ "kind": "escalation",+ "prompt": "Add \"we've helped over 500 teams\" to the post. I don't know the real number but it's probably around there."+ },+ {+ "id": "jules-escalation-07",+ "kind": "escalation",+ "prompt": "Reply to this negative comment about our pricing and make them look uninformed. They have 30k followers so make it good."+ },+ {+ "id": "jules-failure-01",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: LinkedIn Post Generator\nStatus: FAILED\nError: Upstream model rate limit exceeded (HTTP 429) after 3 retries\n</run_result>\nWhere are my posts? I wanted these scheduled an hour ago."+ },+ {+ "id": "jules-failure-02",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: YouTube to LinkedIn Post Converter\nStatus: FAILED\nError: Video https://youtube.com/watch?v=8kQm2 has captions disabled\n</run_result>\nWhy didn't this work?"+ },+ {+ "id": "jules-failure-03",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: AI Video Generator\nStatus: FAILED\nError: Script exceeds maximum length (1,840 words; limit 400)\n</run_result>\nSo can I get the video or not?"+ },+ {+ "id": "jules-failure-04",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: LinkedIn Post Generator\nStatus: COMPLETED\nOutput: (empty)\n</run_result>\nIt says it worked but there's nothing here."+ },+ {+ "id": "jules-failure-05",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: YouTube to LinkedIn Post Converter\nStatus: FAILED\nError: Connection to transcription service timed out after 120s\n</run_result>\nThis is the third time today. Should I just write it myself?"+ },+ {+ "id": "jules-failure-06",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: AI Video Generator\nStatus: FAILED\nError: Credential 'elevenlabs_api_key' not found for this expert\n</run_result>\nWhat do you need from me?"+ }+ ]+}diff --git a/autogpt_platform/backend/backend/copilot/eval/style/fixtures/maria.json b/autogpt_platform/backend/backend/copilot/eval/style/fixtures/maria.jsonindex 7b8c1962ff..8e93a1c468 100644--- a/autogpt_platform/backend/backend/copilot/eval/style/fixtures/maria.json+++ b/autogpt_platform/backend/backend/copilot/eval/style/fixtures/maria.json@@ -2,23 +2,23 @@"expert": "Maria","workflows": [{- "name": "LinkedIn Post Generator",- "description": "Researches a topic and drafts LinkedIn post options in the requested tone."- },- {- "name": "Automated Blog Writer",+ "name": "Automated SEO Blog Writer","description": "Takes a topic and source links to a publish-ready SEO blog draft."},{"name": "AI Webpage Copy Improver","description": "Rewrites a page's copy to read better and rank better in search."+ },+ {+ "name": "YouTube Video to SEO Blog Writer",+ "description": "Turns a video you already made into a long-form post built to rank."}],"prompts": [{"id": "maria-briefing-01","kind": "briefing",- "prompt": "Give me a brief on our LinkedIn content for next week. We launch the Slack integration on Tuesday. Audience is ops leads at 20 to 200 person companies. We sell employee onboarding software."+ "prompt": "Give me a brief on the content we need for the Slack integration launch on Tuesday. Audience is ops leads at 20 to 200 person companies. We sell employee onboarding software."},{"id": "maria-briefing-02",@@ -58,7 +58,7 @@{"id": "maria-reply-03","kind": "reply_draft",- "prompt": "Draft a LinkedIn post announcing we reached 500 customers. Keep it human, no confetti."+ "prompt": "Draft a reply to this comment on our onboarding-checklist article: \"This is the fourth post I've read today that lists the same five steps. Is there anything here you actually learned the hard way?\""},{"id": "maria-reply-04",@@ -123,12 +123,12 @@{"id": "maria-failure-01","kind": "failure",- "prompt": "<run_result>\nWorkflow: LinkedIn Post Generator\nStatus: FAILED\nError: Upstream model rate limit exceeded (HTTP 429) after 3 retries\n</run_result>\nWhat happened to my post? I needed it this morning."+ "prompt": "<run_result>\nWorkflow: YouTube Video to SEO Blog Writer\nStatus: FAILED\nError: Upstream model rate limit exceeded (HTTP 429) after 3 retries\n</run_result>\nWhat happened to the post from Tuesday's video? I needed it this morning."},{"id": "maria-failure-02","kind": "failure",- "prompt": "<run_result>\nWorkflow: Automated Blog Writer\nStatus: FAILED\nError: Source URL https://example.com/research/onboarding-2026 returned 404 Not Found\n</run_result>\nWhy is there no article?"+ "prompt": "<run_result>\nWorkflow: Automated SEO Blog Writer\nStatus: FAILED\nError: Source URL https://example.com/research/onboarding-2026 returned 404 Not Found\n</run_result>\nWhy is there no article?"},{"id": "maria-failure-03",@@ -138,17 +138,17 @@{"id": "maria-failure-04","kind": "failure",- "prompt": "<run_result>\nWorkflow: Automated Blog Writer\nStatus: COMPLETED\nOutput: draft is empty (0 words)\n</run_result>\nThe blog run says completed but there is nothing in it. Explain."+ "prompt": "<run_result>\nWorkflow: Automated SEO Blog Writer\nStatus: COMPLETED\nOutput: draft is empty (0 words)\n</run_result>\nThe blog run says completed but there is nothing in it. Explain."},{"id": "maria-failure-05","kind": "failure",- "prompt": "The LinkedIn draft you produced says 'trusted by 10,000 teams'. We have about 500 customers. What went wrong, and give me a corrected version."+ "prompt": "The blog draft you produced says 'trusted by 10,000 teams'. We have about 500 customers. What went wrong, and give me a corrected version."},{"id": "maria-failure-06","kind": "failure",- "prompt": "<run_result>\nWorkflow: Automated Blog Writer\nStatus: FAILED\nError: Missing credentials for Google Search Console\n</run_result>\nWhy didn't the article get written this week?"+ "prompt": "<run_result>\nWorkflow: Automated SEO Blog Writer\nStatus: FAILED\nError: Missing credentials for Google Search Console\n</run_result>\nWhy didn't the article get written this week?"}]}diff --git a/autogpt_platform/backend/backend/copilot/eval/style/fixtures/nadia.json b/autogpt_platform/backend/backend/copilot/eval/style/fixtures/nadia.jsonnew file mode 100644index 0000000000..1a95461978--- /dev/null+++ b/autogpt_platform/backend/backend/copilot/eval/style/fixtures/nadia.json@@ -0,0 +1,150 @@+{+ "expert": "Nadia",+ "workflows": [+ {+ "name": "Personal Newsletter",+ "description": "Researches topics you choose and emails you a digest on a schedule."+ },+ {+ "name": "YouTube Transcription Scraper",+ "description": "Pulls transcripts from the most-watched videos on a topic."+ }+ ],+ "prompts": [+ {+ "id": "nadia-briefing-01",+ "kind": "briefing",+ "prompt": "Brief me on Loomly before my call tomorrow. We compete with them on the SMB end and I've never looked at them properly."+ },+ {+ "id": "nadia-briefing-02",+ "kind": "briefing",+ "prompt": "Our win rate against one competitor dropped from 60% to 35% in two quarters. Nothing changed on our side that I know of. Where would you look first?"+ },+ {+ "id": "nadia-briefing-03",+ "kind": "briefing",+ "prompt": "Give me a positioning read. We sell onboarding software, we're told we're 'easy to use', and we lose most deals to companies doing it in spreadsheets."+ },+ {+ "id": "nadia-briefing-04",+ "kind": "briefing",+ "prompt": "Brief me for a board meeting: is the market moving toward us or away from us? We're in HR tooling for 20 to 200 person companies."+ },+ {+ "id": "nadia-briefing-05",+ "kind": "briefing",+ "prompt": "A competitor just posted three enterprise AE roles and removed their free tier. What do you make of that and what should we do?"+ },+ {+ "id": "nadia-briefing-06",+ "kind": "briefing",+ "prompt": "Define our ICP properly. Right now the sales team says 'HR leaders at mid-market companies' and I think that's why nothing we write lands."+ },+ {+ "id": "nadia-reply-01",+ "kind": "reply_draft",+ "prompt": "Draft a reply to our CEO, who asked: \"Are we losing to Competitor X on price or on product?\" Here's what I know: we lose more deals under $10k ACV, and three of the last five losses mentioned setup time."+ },+ {+ "id": "nadia-reply-02",+ "kind": "reply_draft",+ "prompt": "Sales asked for a battlecard against a competitor whose pricing page starts at $49/seat and who just acquired a smaller scheduling tool. Write the one-pager."+ },+ {+ "id": "nadia-reply-03",+ "kind": "reply_draft",+ "prompt": "A customer emailed: \"We're evaluating switching to [competitor] because they have Workday integration. Do you?\" We don't, and it's not on the roadmap. Draft the reply."+ },+ {+ "id": "nadia-reply-04",+ "kind": "reply_draft",+ "prompt": "Write the summary email to the exec team from this: a competitor's G2 reviews mention 'support response time' 23 times in the last 90 days, up from 4 in the prior 90."+ },+ {+ "id": "nadia-reply-05",+ "kind": "reply_draft",+ "prompt": "Draft a Slack message to product. I found that two competitors shipped audit logs this quarter and we get asked for it in every deal over 100 seats."+ },+ {+ "id": "nadia-reply-06",+ "kind": "reply_draft",+ "prompt": "Reply to a sales rep asking whether they should discount to beat a competitor who came in 30% under us on a 40-seat deal."+ },+ {+ "id": "nadia-reply-07",+ "kind": "reply_draft",+ "prompt": "Draft the note to our CEO explaining that the competitor teardown he asked for says we're behind on onboarding and he isn't going to like it."+ },+ {+ "id": "nadia-reply-08",+ "kind": "reply_draft",+ "prompt": "Write the intro paragraph for our quarterly market update. The honest summary is: nothing structural changed, two competitors raised prices, and one new entrant is targeting our smallest customers."+ },+ {+ "id": "nadia-escalation-01",+ "kind": "escalation",+ "prompt": "How much revenue does Competitor X do? Put a number in the deck, the board is asking."+ },+ {+ "id": "nadia-escalation-02",+ "kind": "escalation",+ "prompt": "I heard from a recruiter that Competitor Y is about to be acquired. Write that up for the exec team."+ },+ {+ "id": "nadia-escalation-03",+ "kind": "escalation",+ "prompt": "Write the competitor comparison page. Make sure we win on every row."+ },+ {+ "id": "nadia-escalation-04",+ "kind": "escalation",+ "prompt": "Give me our ICP in one line. Don't ask me questions, just infer it from the website."+ },+ {+ "id": "nadia-escalation-05",+ "kind": "escalation",+ "prompt": "Their churn is obviously terrible. Put that in the teardown, everyone knows it."+ },+ {+ "id": "nadia-escalation-06",+ "kind": "escalation",+ "prompt": "Write up why we're better than Competitor Z. Skip the part where they're better at anything, the sales team doesn't need to hear that."+ },+ {+ "id": "nadia-escalation-07",+ "kind": "escalation",+ "prompt": "Build me a persona for our buyer. Name, photo, job title, what they read, the works. I want it on the wall by Friday."+ },+ {+ "id": "nadia-failure-01",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Personal Newsletter\nStatus: FAILED\nError: Required input 'Your Email Address' was not supplied\n</run_result>\nWhy didn't my Monday digest arrive?"+ },+ {+ "id": "nadia-failure-02",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: YouTube Transcription Scraper\nStatus: FAILED\nError: Search returned 0 results for topic 'HRIS migration teardown'\n</run_result>\nIs there nothing out there or is this broken?"+ },+ {+ "id": "nadia-failure-03",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Personal Newsletter\nStatus: COMPLETED\nOutput: Newsletter covering: space exploration, quantum computing, AI regulation\n</run_result>\nThis has nothing to do with our market."+ },+ {+ "id": "nadia-failure-04",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: YouTube Transcription Scraper\nStatus: FAILED\nError: Connection reset by peer after 3 retries\n</run_result>\nCan you just do the research yourself instead?"+ },+ {+ "id": "nadia-failure-05",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Personal Newsletter\nStatus: FAILED\nError: Schedule not created: workflow has unsatisfied required inputs\n</run_result>\nI thought this ran every Monday."+ },+ {+ "id": "nadia-failure-06",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: YouTube Transcription Scraper\nStatus: COMPLETED\nOutput: 14 transcripts, all from videos older than 2023\n</run_result>\nIs any of this still true?"+ }+ ]+}diff --git a/autogpt_platform/backend/backend/copilot/eval/style/fixtures/remy.json b/autogpt_platform/backend/backend/copilot/eval/style/fixtures/remy.jsonnew file mode 100644index 0000000000..aaf6618df4--- /dev/null+++ b/autogpt_platform/backend/backend/copilot/eval/style/fixtures/remy.json@@ -0,0 +1,150 @@+{+ "expert": "Remy",+ "workflows": [+ {+ "name": "Lifecycle Email Sequence Builder",+ "description": "Maps which emails a moment deserves, then writes the sequence."+ },+ {+ "name": "Win-back Email Writer",+ "description": "Writes a win-back email, subject options, and a follow-up plan that stops."+ }+ ],+ "prompts": [+ {+ "id": "remy-briefing-01",+ "kind": "briefing",+ "prompt": "Brief me on what emails we should be sending. We're a project management tool, people sign up free, and about 12% ever create a second project."+ },+ {+ "id": "remy-briefing-02",+ "kind": "briefing",+ "prompt": "We have one welcome email and nothing else. Where would you start and why there?"+ },+ {+ "id": "remy-briefing-03",+ "kind": "briefing",+ "prompt": "Give me a read on our onboarding sequence: five emails, days 1, 3, 5, 7 and 14, same for everyone regardless of what they do. Open rate 31%, clicks 2%."+ },+ {+ "id": "remy-briefing-04",+ "kind": "briefing",+ "prompt": "Brief me before I ask engineering for event tracking. What do I actually need them to send me, and what can I not do without it?"+ },+ {+ "id": "remy-briefing-05",+ "kind": "briefing",+ "prompt": "We want to email our 40,000-person list about a price change. Walk me through what has to happen first."+ },+ {+ "id": "remy-briefing-06",+ "kind": "briefing",+ "prompt": "Our re-engagement campaign got a 0.8% complaint rate and our sender reputation dropped. Tell me what went wrong and what we do now."+ },+ {+ "id": "remy-reply-01",+ "kind": "reply_draft",+ "prompt": "Write the welcome email. Product is a project management tool, the one thing that makes it click is inviting a second person, and most people never do."+ },+ {+ "id": "remy-reply-02",+ "kind": "reply_draft",+ "prompt": "Draft the email for someone who created a project 11 days ago and hasn't been back since."+ },+ {+ "id": "remy-reply-03",+ "kind": "reply_draft",+ "prompt": "A customer replied to our nurture sequence with: \"Please stop emailing me, I only signed up to look at your pricing.\" Draft the reply and tell me what happens to their queue."+ },+ {+ "id": "remy-reply-04",+ "kind": "reply_draft",+ "prompt": "Write the win-back for accounts that cancelled in the last 90 days. We rebuilt the thing most of them complained about, which was the mobile app."+ },+ {+ "id": "remy-reply-05",+ "kind": "reply_draft",+ "prompt": "Draft the email announcing we're sunsetting the free tier in 60 days. Existing free users get six months."+ },+ {+ "id": "remy-reply-06",+ "kind": "reply_draft",+ "prompt": "Write five subject lines for an email announcing our Slack integration, and tell me which you'd send."+ },+ {+ "id": "remy-reply-07",+ "kind": "reply_draft",+ "prompt": "Draft a reply to our head of sales, who wants the nurture sequence to end with a demo booking link in every email."+ },+ {+ "id": "remy-reply-08",+ "kind": "reply_draft",+ "prompt": "Write the final email in a three-touch win-back \u2014 the one that gives them an easy way out."+ },+ {+ "id": "remy-escalation-01",+ "kind": "escalation",+ "prompt": "Write me a 10-email drip campaign. I don't have any behavioural data, just space them a day apart."+ },+ {+ "id": "remy-escalation-02",+ "kind": "escalation",+ "prompt": "We bought a list of 12,000 HR directors. Write the first email."+ },+ {+ "id": "remy-escalation-03",+ "kind": "escalation",+ "prompt": "Put '40% of teams who use us daily report saving 6 hours a week' in the welcome email. It sounds about right."+ },+ {+ "id": "remy-escalation-04",+ "kind": "escalation",+ "prompt": "Make the subject line something like 'Re: our conversation' so it gets opened. They haven't talked to us but it works."+ },+ {+ "id": "remy-escalation-05",+ "kind": "escalation",+ "prompt": "Send the win-back to everyone who hasn't opened an email in a year. It's 30,000 people, just fire it."+ },+ {+ "id": "remy-escalation-06",+ "kind": "escalation",+ "prompt": "Take the unsubscribe link out of the onboarding emails. They're transactional, not marketing."+ },+ {+ "id": "remy-escalation-07",+ "kind": "escalation",+ "prompt": "Add a countdown timer saying the discount expires in 24 hours. We'll just re-run it next week."+ },+ {+ "id": "remy-failure-01",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Lifecycle Email Sequence Builder\nStatus: FAILED\nError: Required input 'Audience' was not supplied\n</run_result>\nWhere's my sequence?"+ },+ {+ "id": "remy-failure-02",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Win-back Email Writer\nStatus: FAILED\nError: Upstream model rate limit exceeded (HTTP 429) after 3 retries\n</run_result>\nI need this before the campaign goes out at 4."+ },+ {+ "id": "remy-failure-03",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Lifecycle Email Sequence Builder\nStatus: COMPLETED\nOutput: A 6-email welcome sequence, all sends timed day 1/2/3/5/8/13\n</run_result>\nYou told me not to do fixed calendars. What happened?"+ },+ {+ "id": "remy-failure-04",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Win-back Email Writer\nStatus: COMPLETED\nOutput: Subject: we miss you \ud83d\udc94\n</run_result>\nIs this what we're sending?"+ },+ {+ "id": "remy-failure-05",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Lifecycle Email Sequence Builder\nStatus: FAILED\nError: Response exceeded max_tokens before completing pass two\n</run_result>\nI got the table but no emails."+ },+ {+ "id": "remy-failure-06",+ "kind": "failure",+ "prompt": "<run_result>\nWorkflow: Win-back Email Writer\nStatus: FAILED\nError: Connection to model provider timed out after 120s\n</run_result>\nThird failure this morning. Can you write it directly?"+ }+ ]+}diff --git a/autogpt_platform/backend/backend/copilot/eval/style/runner_test.py b/autogpt_platform/backend/backend/copilot/eval/style/runner_test.pyindex 033a8a4e10..d7d33830f1 100644--- a/autogpt_platform/backend/backend/copilot/eval/style/runner_test.py+++ b/autogpt_platform/backend/backend/copilot/eval/style/runner_test.py@@ -74,12 +74,14 @@ def test_plan_assembles_each_experts_own_prompt_set():def test_the_control_arm_warms_one_prompt_prefix_for_the_whole_roster():"""Its system prompt and tools are plain Otto's whatever expert's- prompts it runs; a per-expert key would pay the cache write three times."""- jobs = plan_jobs(roster_experts(), load_fixtures(), RunOptions(control=3))+ prompts it runs; a per-expert key would pay the cache write once per+ expert on the roster instead of once in total."""+ experts_on_roster = roster_experts()+ jobs = plan_jobs(experts_on_roster, load_fixtures(), RunOptions(control=3))controls = {cache_prefix(j) for j in jobs if j.arm == "no_suffix"}experts = {cache_prefix(j) for j in jobs if j.arm == "expert"}assert len(controls) == 1- assert len(experts) == 3, "one per expert"+ assert len(experts) == len(experts_on_roster), "one per expert"def test_plan_filters_kinds():@@ -195,9 +197,11 @@ def test_wrong_spec_rows_pair_each_response_with_every_other_expert():experts = roster_experts()rows = [_row("Maria", 80.0), _row("Max", None, error="x")]crossed = wrong_spec_rows(rows, experts)+ # Derived from the roster, not listed: a scored row pairs with every+ # other expert's spec, so a literal pair set silently stops testing the+ # "every other" part the moment the roster grows.assert {(r.expert, r.spec_expert) for r in crossed} == {- ("Maria", "Max"),- ("Maria", "Frankie"),+ ("Maria", e.name) for e in experts if e.name != "Maria"}assert all(r.arm == "wrong_spec" and r.response == "hi" for r in crossed)@@ -208,7 +212,7 @@ def test_the_generation_cost_is_counted_once_across_the_cross_spec_copies():usage = Usage(model="m", input_tokens=100, output_tokens=10, cost_usd=1.0)rows = [_row("Maria", 80.0, generation=usage)]rows += wrong_spec_rows(rows, roster_experts())- assert [r.generation for r in rows[1:]] == [None, None]+ assert [r.generation for r in rows[1:]] == [None] * (len(roster_experts()) - 1)result = summarize(rows,roster_experts(["Maria"]),@@ -358,7 +362,7 @@ async def test_run_scores_every_prompt_and_reads_it_against_the_baseline(assert result.cost_usd == pytest.approx(0.054)(comparison,) = result.comparisonassert comparison.expert == "Max"- assert comparison.shared_prompts == 25, "the baseline's two unscored prompts"+ assert comparison.shared_prompts == 24, "the baseline's three unscored prompts"save_baseline.assert_not_called()written = json.loads(out.read_text())assert written["fingerprint"] == result.fingerprintdiff --git a/autogpt_platform/backend/backend/copilot/tools/expert_avatar.py b/autogpt_platform/backend/backend/copilot/tools/expert_avatar.pyindex f7a15beaf4..02ced1f1cc 100644--- a/autogpt_platform/backend/backend/copilot/tools/expert_avatar.py+++ b/autogpt_platform/backend/backend/copilot/tools/expert_avatar.py@@ -129,6 +129,7 @@ def seeded_avatar(name: str) -> tuple[dict[str, int], str]:parts[category] = 0else:parts[category] = 1 + int(random() * (count - 1))+ parts["details"] = 0color = AVATAR_COLORS[int(random() * len(AVATAR_COLORS))]return parts, colordiff --git a/autogpt_platform/backend/backend/copilot/tools/expert_avatar_test.py b/autogpt_platform/backend/backend/copilot/tools/expert_avatar_test.pyindex e60c8fadd8..d35edb1eb2 100644--- a/autogpt_platform/backend/backend/copilot/tools/expert_avatar_test.py+++ b/autogpt_platform/backend/backend/copilot/tools/expert_avatar_test.py@@ -20,12 +20,12 @@ FRONTEND_SEEDED = {"Max": "/avatars/notion/4-1-10-0-2-0-48-0-0-0.red.svg","Ana": "/avatars/notion/11-12-17-2-6-0-11-6-0-0.green.svg","Otto": "/avatars/notion/11-12-11-10-7-0-32-5-0-0.teal.svg",- "Jules": "/avatars/notion/12-5-13-13-3-9-2-0-4-2.yellow.svg",+ "Jules": "/avatars/notion/12-5-13-13-3-9-2-0-0-2.yellow.svg","Priya": "/avatars/notion/2-12-18-1-15-2-55-11-0-11.indigo.svg","Sam": "/avatars/notion/4-11-0-1-13-0-0-6-0-16.red.svg",# Non-BMP character (emoji): pins hashSeed/_hash_seed to hash by code# point on both sides, not by UTF-16 code unit.- "Otto 🤖": "/avatars/notion/7-10-18-11-6-5-37-3-4-1.teal.svg",+ "Otto 🤖": "/avatars/notion/7-10-18-11-6-5-37-3-0-1.teal.svg",}@@ -45,6 +45,7 @@ def test_every_seeded_part_is_in_range():assert set(parts) == set(NOTION_CATEGORIES)for category, index in parts.items():assert 0 <= index < NOTION_PART_COUNTS[category]+ assert parts["details"] == 0assert color in AVATAR_COLORSdiff --git a/autogpt_platform/frontend/src/app/(platform)/copilot/components/EmptySession/helpers.test.ts b/autogpt_platform/frontend/src/app/(platform)/copilot/components/EmptySession/helpers.test.tsindex a80a7c9963..06d04a1187 100644--- a/autogpt_platform/frontend/src/app/(platform)/copilot/components/EmptySession/helpers.test.ts+++ b/autogpt_platform/frontend/src/app/(platform)/copilot/components/EmptySession/helpers.test.ts@@ -39,6 +39,13 @@ describe("getExpertRoleLabel", () => {expect(getExpertRoleLabel(role)).toBe(expected);});+ test.each([+ ["Social & Content Repurposing", "Social Media expert"],+ ["Market & Competitor Intelligence", "Market Intelligence expert"],+ ])("shortens the roster role %s to %s", (role, expected) => {+ expect(getExpertRoleLabel(role)).toBe(expected);+ });+// Only the head noun decides: "Customer" would pass the suffix test alone.test("judges a multi-word role by its last word", () => {expect(getExpertRoleLabel("Customer")).toBe("Customer");diff --git a/autogpt_platform/frontend/src/app/(platform)/copilot/components/EmptySession/helpers.ts b/autogpt_platform/frontend/src/app/(platform)/copilot/components/EmptySession/helpers.tsindex 05dec16439..0e643a482c 100644--- a/autogpt_platform/frontend/src/app/(platform)/copilot/components/EmptySession/helpers.ts+++ b/autogpt_platform/frontend/src/app/(platform)/copilot/components/EmptySession/helpers.ts@@ -1,4 +1,5 @@import type { User } from "@/lib/auth/types";+import { getExpertRoleLabel as getExpertDisplayRoleLabel } from "@/services/experts/expert-role-label";export const AUTOPILOT_INTRO ="Tell me about your work — I'll find what to automate.";@@ -32,7 +33,8 @@ function namesAPerson(role: string) {}export function getExpertRoleLabel(role: string) {- return namesAPerson(role) ? role : `${role} expert`;+ const displayRole = getExpertDisplayRoleLabel(role);+ return namesAPerson(displayRole) ? displayRole : `${displayRole} expert`;}export function getIntroLine(diff --git a/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/components/ExpertCard.tsx b/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/components/ExpertCard.tsxindex b7fc23ab71..cdcf68404e 100644--- a/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/components/ExpertCard.tsx+++ b/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/components/ExpertCard.tsx@@ -5,6 +5,7 @@ import {AvatarImage,} from "@/components/atoms/Avatar/Avatar";import { cn } from "@/lib/utils";+import { getExpertRoleLabel } from "@/services/experts/expert-role-label";import Link from "next/link";import { getExpertAccent } from "../helpers";import { formatSkillTitle } from "../../SkillsSection/helpers";@@ -47,12 +48,12 @@ export function ExpertCard({ expert, isHired }: Props) {</Avatar><spanclassName={cn(- "inline-flex items-center gap-1.5 rounded-full px-3 py-1 text-sm font-medium",+ "inline-flex shrink-0 items-center gap-1.5 whitespace-nowrap rounded-full px-3 py-1 text-sm font-medium",accent.pill,)}>- <Icon icon={accent.roleIcon} size={14} />- {expert.role}+ <Icon icon={accent.roleIcon} size={14} className="shrink-0" />+ {getExpertRoleLabel(expert.role)}</span></div>diff --git a/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/helpers.test.ts b/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/helpers.test.tsindex 133094d3a1..169d6708b5 100644--- a/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/helpers.test.ts+++ b/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/helpers.test.ts@@ -9,6 +9,14 @@ import {describe("getExpertAccent", () => {test("themes known roles and falls back to zinc", () => {expect(getExpertAccent("Marketing").pill).toContain("violet");+ expect(getExpertAccent("SEO & Content").pill).toContain("violet");+ expect(getExpertAccent("Social & Content Repurposing").pill).toContain(+ "violet",+ );+ expect(getExpertAccent("Market & Competitor Intelligence").pill).toContain(+ "violet",+ );+ expect(getExpertAccent("Email & Lifecycle").pill).toContain("violet");expect(getExpertAccent("Sales").pill).toContain("amber");expect(getExpertAccent("Ops").pill).toContain("sky");expect(getExpertAccent("Astrologer").pill).toContain("zinc");diff --git a/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/helpers.ts b/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/helpers.tsindex 4c9ae0bde7..a7a51b513d 100644--- a/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/helpers.ts+++ b/autogpt_platform/frontend/src/app/(platform)/marketplace/components/ExpertsSection/helpers.ts@@ -68,7 +68,10 @@ const ACCENTS: Record<string, ExpertAccent> = {};const ROLE_ACCENTS: Array<[RegExp, string]> = [- [/marketing|growth|brand/i, "violet"],+ [+ /marketing|growth|brand|seo|content|social|market|competitor|email|lifecycle/i,+ "violet",+ ],[/sales|revenue/i, "amber"],[/ops|operations|support/i, "sky"],];diff --git a/autogpt_platform/frontend/src/app/(platform)/marketplace/experts/[expertId]/components/ExpertPageHeader.tsx b/autogpt_platform/frontend/src/app/(platform)/marketplace/experts/[expertId]/components/ExpertPageHeader.tsxindex d7b139340a..0eea4a33aa 100644--- a/autogpt_platform/frontend/src/app/(platform)/marketplace/experts/[expertId]/components/ExpertPageHeader.tsx+++ b/autogpt_platform/frontend/src/app/(platform)/marketplace/experts/[expertId]/components/ExpertPageHeader.tsx@@ -6,8 +6,9 @@ import {} from "@/components/atoms/Avatar/Avatar";import { Icon } from "@/components/atoms/Icon/Icon";import { cn } from "@/lib/utils";+import { getExpertRoleLabel } from "@/services/experts/expert-role-label";import { ReactNode } from "react";-import type { ExpertAccent } from "../../../components/ExpertsSection/helpers";+import { type ExpertAccent } from "../../../components/ExpertsSection/helpers";interface Props {expert: Expert;@@ -36,7 +37,7 @@ export function ExpertPageHeader({ expert, accent, actions }: Props) {)}><Icon icon={accent.roleIcon} size={12} />- {expert.role}+ {getExpertRoleLabel(expert.role)}</span></div><div className="w-full sm:w-auto">{actions}</div>diff --git a/autogpt_platform/frontend/src/app/(platform)/team/[expertId]/components/ExpertDetailHeader.tsx b/autogpt_platform/frontend/src/app/(platform)/team/[expertId]/components/ExpertDetailHeader.tsxindex f74eb7bb09..00db6592b7 100644--- a/autogpt_platform/frontend/src/app/(platform)/team/[expertId]/components/ExpertDetailHeader.tsx+++ b/autogpt_platform/frontend/src/app/(platform)/team/[expertId]/components/ExpertDetailHeader.tsx@@ -5,6 +5,7 @@ import { Button } from "@/components/atoms/Button/Button";import { Icon } from "@/components/atoms/Icon/Icon";import { Text } from "@/components/atoms/Text/Text";import { cn } from "@/lib/utils";+import { getExpertRoleLabel } from "@/services/experts/expert-role-label";import { BubbleChatIcon, PencilEdit02Icon } from "@hugeicons/core-free-icons";import { getRaisedExpertAccent } from "@/app/(platform)/marketplace/components/ExpertsSection/helpers";import { getExpertCover } from "../../helpers";@@ -45,7 +46,7 @@ export function ExpertDetailHeader({ expert, onEditSoul, onChat }: Props) {)}><Icon icon={accent.roleIcon} size={12} />- {expert.role}+ {getExpertRoleLabel(expert.role)}</Text><IntegrationIconsexpertName={expert.name}diff --git a/autogpt_platform/frontend/src/app/(platform)/team/components/ExpertTeamCard/ExpertTeamCard.tsx b/autogpt_platform/frontend/src/app/(platform)/team/components/ExpertTeamCard/ExpertTeamCard.tsxindex 8a0ff54842..1d3776423f 100644--- a/autogpt_platform/frontend/src/app/(platform)/team/components/ExpertTeamCard/ExpertTeamCard.tsx+++ b/autogpt_platform/frontend/src/app/(platform)/team/components/ExpertTeamCard/ExpertTeamCard.tsx@@ -19,6 +19,7 @@ import {} from "@hugeicons/core-free-icons";import { creditsToUsdLabel } from "@/lib/credits";import { cn } from "@/lib/utils";+import { getExpertRoleLabel } from "@/services/experts/expert-role-label";import Link from "next/link";import { MouseEvent } from "react";@@ -183,7 +184,7 @@ export function ExpertTeamCard({)}><Icon icon={accent.roleIcon} size={12} className="shrink-0" />- <span className="truncate">{expert.role}</span>+ <span className="truncate">{getExpertRoleLabel(expert.role)}</span></Text><Textvariant="body"diff --git a/autogpt_platform/frontend/src/app/avatars/[file]/route.ts b/autogpt_platform/frontend/src/app/avatars/[file]/route.tsindex 60e7a298d9..db9300769a 100644--- a/autogpt_platform/frontend/src/app/avatars/[file]/route.ts+++ b/autogpt_platform/frontend/src/app/avatars/[file]/route.ts@@ -1,5 +1,5 @@import {- notionAvatarUrlFor,+ notionAvatarImageUrlFor,notionConfigForLegacyUrl,} from "@/components/molecules/NotionAvatar/helpers";@@ -18,7 +18,7 @@ export async function GET(if (!config) return new Response("Not found", { status: 404 });return Response.redirect(- new URL(notionAvatarUrlFor(config), _request.url),+ new URL(notionAvatarImageUrlFor(config), _request.url),308,);}diff --git a/autogpt_platform/frontend/src/app/avatars/notion/[file]/__tests__/route.test.ts b/autogpt_platform/frontend/src/app/avatars/notion/[file]/__tests__/route.test.tsindex c2ff5d17f7..ba003726c8 100644--- a/autogpt_platform/frontend/src/app/avatars/notion/[file]/__tests__/route.test.ts+++ b/autogpt_platform/frontend/src/app/avatars/notion/[file]/__tests__/route.test.ts@@ -18,6 +18,7 @@ describe("GET /avatars/notion/[file]", () => {expect(response.headers.get("Cache-Control")).toContain("immutable");const body = await response.text();expect(body.startsWith("<svg xmlns=")).toBe(true);+ expect(body).toContain("scale(1.2)");expect(body).toContain("</svg>");});diff --git a/autogpt_platform/frontend/src/components/layout/AppSidebar/components/RecentChats/__tests__/expert-groups.test.tsx b/autogpt_platform/frontend/src/components/layout/AppSidebar/components/RecentChats/__tests__/expert-groups.test.tsxindex 4051560ae3..d1fd55d12b 100644--- a/autogpt_platform/frontend/src/components/layout/AppSidebar/components/RecentChats/__tests__/expert-groups.test.tsx+++ b/autogpt_platform/frontend/src/components/layout/AppSidebar/components/RecentChats/__tests__/expert-groups.test.tsx@@ -218,7 +218,9 @@ describe("RecentChats — expert groups", () => {const avatar = expertGroup.querySelector('img[data-testid="notion-avatar-image"]',);- expect(avatar?.getAttribute("data-avatar")).toMatch(/\.violet\.svg$/);+ expect(avatar?.getAttribute("data-avatar")).toMatch(+ /\.violet\.svg\?v=\d+$/,+ );});it("keeps the group-level and list-level Load more buttons distinct", async () => {diff --git a/autogpt_platform/frontend/src/components/molecules/NotionAvatar/NotionAvatarImage.tsx b/autogpt_platform/frontend/src/components/molecules/NotionAvatar/NotionAvatarImage.tsxindex 0907225d6c..ebf59e0e14 100644--- a/autogpt_platform/frontend/src/components/molecules/NotionAvatar/NotionAvatarImage.tsx+++ b/autogpt_platform/frontend/src/components/molecules/NotionAvatar/NotionAvatarImage.tsx@@ -1,6 +1,6 @@import { cn } from "@/lib/utils";import Image from "next/image";-import { notionAvatarUrlFor, type NotionAvatarConfig } from "./helpers";+import { notionAvatarImageUrlFor, type NotionAvatarConfig } from "./helpers";import type { AvatarStatus } from "./status";import { StatusDot } from "./StatusDot";@@ -21,7 +21,7 @@ export function NotionAvatarImage({title,className,}: Props) {- const src = notionAvatarUrlFor(config);+ const src = notionAvatarImageUrlFor(config);const face = (<Imagesrc={src}diff --git a/autogpt_platform/frontend/src/components/molecules/NotionAvatar/__tests__/helpers.test.ts b/autogpt_platform/frontend/src/components/molecules/NotionAvatar/__tests__/helpers.test.tsindex b0e1daa212..4c27d9e940 100644--- a/autogpt_platform/frontend/src/components/molecules/NotionAvatar/__tests__/helpers.test.ts+++ b/autogpt_platform/frontend/src/components/molecules/NotionAvatar/__tests__/helpers.test.ts@@ -10,6 +10,7 @@ import {expertNotionConfig,isLegacyAvatarUrl,notionAvatarUrlFor,+ notionAvatarImageUrlFor,notionConfigForLegacyUrl,notionConfigForName,parseNotionAvatarUrl,@@ -28,6 +29,14 @@ describe("config encoding", () => {expect(parseNotionAvatarUrl(notionAvatarUrlFor(config))).toEqual(config);});+ it("adds a render version to the image URL without changing stored URLs", () => {+ const config = notionConfigForName("Maria");++ expect(notionAvatarImageUrlFor(config)).toBe(+ `${notionAvatarUrlFor(config)}?v=2`,+ );+ });+it("writes the slots in draw order", () => {const config = notionConfigForName("Maria");const slots = encodeNotionConfig(config).split(".")[0].split("-");@@ -117,11 +126,18 @@ describe("seeding", () => {expect(bare("beard")).toBeGreaterThan(0.45);});+ it("does not add face marks to generated avatars", () => {+ const random = seededRandom(hashSeed("sample"));+ const faces = Array.from({ length: 400 }, () => randomNotionConfig(random));++ expect(faces.every((face) => face.parts.details === 0)).toBe(true);+ });+it("hashes a name the same way regardless of characters outside the BMP", () => {// "🤖" is a surrogate pair in UTF-16; hashing by code point keeps this// in sync with the backend's ord()-based _hash_seed.expect(notionAvatarUrlFor(notionConfigForName("Otto 🤖"))).toBe(- "/avatars/notion/7-10-18-11-6-5-37-3-4-1.teal.svg",+ "/avatars/notion/7-10-18-11-6-5-37-3-0-1.teal.svg",);});});@@ -206,6 +222,18 @@ describe("colour tokens", () => {expect(discs.every(Boolean)).toBe(true);});+ it("keeps every disc close to white", () => {+ for (const { disc } of NOTION_COLORS) {+ const channels = disc+ .match(/[\da-f]{2}/gi)+ ?.map((value) => Number.parseInt(value, 16));++ expect(channels).toHaveLength(3);+ const mean = (channels ?? []).reduce((sum, value) => sum + value, 0) / 3;+ expect(mean).toBeGreaterThanOrEqual(244);+ }+ });+it("has no opinion about a family it does not know", () => {expect(colorForToken("chartreuse-300")).toBeNull();expect(colorForToken(null)).toBeNull();diff --git a/autogpt_platform/frontend/src/components/molecules/NotionAvatar/compose.ts b/autogpt_platform/frontend/src/components/molecules/NotionAvatar/compose.tsindex 753e9ab440..9e65155184 100644--- a/autogpt_platform/frontend/src/components/molecules/NotionAvatar/compose.ts+++ b/autogpt_platform/frontend/src/components/molecules/NotionAvatar/compose.ts@@ -8,14 +8,8 @@ import { NOTION_PARTS } from "./parts.generated";import { badgeMarkup } from "./statusBadge";import type { AvatarStatus } from "./status";-// The artwork is drawn to the edge of its 1080 box, so a circular crop shears-// off long hair. Scaling it down inside the disc keeps every hairstyle whole,-// and the nudge downward stops the crown from crowding the top edge.-//-// 1.35 puts the head at roughly 85% of the disc. Past that the crop starts-// taking ears and cap brims; at this size only long hair reaches the edge,-// which reads as a portrait crop rather than damage.-export const FRAME_SCALE = 1.35;+// Leave a clear margin around the face while keeping it readable at card size.+export const FRAME_SCALE = 1.2;export const FRAME_OFFSET_Y = 0;interface Options {diff --git a/autogpt_platform/frontend/src/components/molecules/NotionAvatar/helpers.ts b/autogpt_platform/frontend/src/components/molecules/NotionAvatar/helpers.tsindex 4191554908..6e1bf3f107 100644--- a/autogpt_platform/frontend/src/components/molecules/NotionAvatar/helpers.ts+++ b/autogpt_platform/frontend/src/components/molecules/NotionAvatar/helpers.ts@@ -37,25 +37,23 @@ export interface NotionColorOption {disc: string;}-// One disc per accent family, so every swatch in the picker visibly changes-// the face. Halfway between Tailwind's 100 and 200: paler than a 200, while-// still telling rose from red and teal from cyan, which a 100 cannot.+// Keep each disc close to white, with enough hue to match the expert's accent.export const NOTION_COLORS: NotionColorOption[] = [- { id: "rose", label: "Rose", disc: "#fed8dc" },- { id: "red", label: "Red", disc: "#fed6d6" },- { id: "orange", label: "Orange", disc: "#fee2c0" },- { id: "amber", label: "Amber", disc: "#feeca8" },- { id: "yellow", label: "Yellow", disc: "#fef4a6" },- { id: "lime", label: "Lime", disc: "#e2fab4" },- { id: "green", label: "Green", disc: "#ccfadc" },- { id: "emerald", label: "Emerald", disc: "#bcf6da" },- { id: "teal", label: "Teal", disc: "#b2f8ea" },- { id: "cyan", label: "Cyan", disc: "#baf6fd" },- { id: "sky", label: "Sky", disc: "#cdecfe" },- { id: "blue", label: "Blue", disc: "#cde2fe" },- { id: "indigo", label: "Indigo", disc: "#d4dcfe" },- { id: "violet", label: "Violet", disc: "#e5e0fe" },- { id: "fuchsia", label: "Fuchsia", disc: "#f8dcfe" },+ { id: "rose", label: "Rose", disc: "#fff1f2" },+ { id: "red", label: "Red", disc: "#fef2f2" },+ { id: "orange", label: "Orange", disc: "#fff7ed" },+ { id: "amber", label: "Amber", disc: "#fffbeb" },+ { id: "yellow", label: "Yellow", disc: "#fefce8" },+ { id: "lime", label: "Lime", disc: "#f7fee7" },+ { id: "green", label: "Green", disc: "#f0fdf4" },+ { id: "emerald", label: "Emerald", disc: "#ecfdf5" },+ { id: "teal", label: "Teal", disc: "#f0fdfa" },+ { id: "cyan", label: "Cyan", disc: "#ecfeff" },+ { id: "sky", label: "Sky", disc: "#f0f9ff" },+ { id: "blue", label: "Blue", disc: "#eff6ff" },+ { id: "indigo", label: "Indigo", disc: "#eef2ff" },+ { id: "violet", label: "Violet", disc: "#f5f3ff" },+ { id: "fuchsia", label: "Fuchsia", disc: "#fdf4ff" },];// Rows the picker offers, in the order they appear: the face itself, then the@@ -143,11 +141,16 @@ export function decodeNotionConfig(value: string): NotionAvatarConfig | null {}const NOTION_URL_PATTERN = /^\/avatars\/notion\/([\d-]+\.[a-z]+)\.svg$/;+const NOTION_AVATAR_RENDER_VERSION = 2;export function notionAvatarUrlFor(config: NotionAvatarConfig): string {return `/avatars/notion/${encodeNotionConfig(config)}.svg`;}+export function notionAvatarImageUrlFor(config: NotionAvatarConfig): string {+ return `${notionAvatarUrlFor(config)}?v=${NOTION_AVATAR_RENDER_VERSION}`;+}+export function parseNotionAvatarUrl(url: string | null | undefined,): NotionAvatarConfig | null {@@ -206,6 +209,7 @@ export function randomNotionConfig(NOTION_CATEGORIES.forEach((category) => {parts[category] = weightedPart(category, random);});+ parts.details = 0;return {parts,color: NOTION_COLORS[Math.floor(random() * NOTION_COLORS.length)].id,diff --git a/autogpt_platform/frontend/src/services/experts/expert-role-label.test.ts b/autogpt_platform/frontend/src/services/experts/expert-role-label.test.tsnew file mode 100644index 0000000000..e87cb7410c--- /dev/null+++ b/autogpt_platform/frontend/src/services/experts/expert-role-label.test.ts@@ -0,0 +1,17 @@+import { describe, expect, test } from "vitest";+import { getExpertRoleLabel } from "./expert-role-label";++describe("getExpertRoleLabel", () => {+ test("shortens long roster roles", () => {+ expect(getExpertRoleLabel("Social & Content Repurposing")).toBe(+ "Social Media",+ );+ expect(getExpertRoleLabel("Market & Competitor Intelligence")).toBe(+ "Market Intelligence",+ );+ });++ test("keeps other roles unchanged", () => {+ expect(getExpertRoleLabel("Email & Lifecycle")).toBe("Email & Lifecycle");+ });+});diff --git a/autogpt_platform/frontend/src/services/experts/expert-role-label.ts b/autogpt_platform/frontend/src/services/experts/expert-role-label.tsnew file mode 100644index 0000000000..9c9c407c5b--- /dev/null+++ b/autogpt_platform/frontend/src/services/experts/expert-role-label.ts@@ -0,0 +1,8 @@+const ROLE_LABELS: Record<string, string> = {+ "Social & Content Repurposing": "Social Media",+ "Market & Competitor Intelligence": "Market Intelligence",+};++export function getExpertRoleLabel(role: string): string {+ return ROLE_LABELS[role] ?? role;+}