Skip to content

graff — a fast agentic coding harness in Zig: multi-provider, MCP, workflows, DGM evolution loop, TS/Python SDKs

Python

feat(provider): add oMLX as a built-in local provider

Rach Pradhanc10c826fc3b7
diff --git a/CHANGELOG.md b/CHANGELOG.md
index c191af42..e9a1564b 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -10,6 +10,12 @@ The release workflow uses a tag's section here as its release notes (a
hand-written `docs/releases/<tag>.md` wins if present), so keeping this file
current is part of cutting a release.
+## Unreleased
+
+- oMLX is a built-in local provider (`omlx`, `127.0.0.1:8000`). Set its key with `graff key set omlx <key>` or `OMLX_API_KEY`. `graff models` lists the models it serves, `--model omlx` starts on the first one, `/effort` is sent as `reasoning_effort`, and it is billed as local.
+- Live model catalogs read a `max_model_len` context window, which oMLX and vLLM-style servers report.
+- When a local server refuses the connection, the failed turn names its URL, so a stopped server is easy to spot.
+
## v0.302.47
- When a compaction can't get the context under the compaction point, the next one waits for real growth instead of running on every step (#1686).
diff --git a/README.md b/README.md
index 067bdb10..f518037b 100644
--- a/README.md
+++ b/README.md
@@ -220,6 +220,13 @@ Vercel, OpenRouter, MiniMax, Xiaomi, Groq, Cerebras, Mistral, plus one
workspace router in `.graff/.config.router`. `graff models refresh` pulls the
catalogs. Claude-subscription OAuth is deliberately not supported.
+Local servers: `omlx` (the oMLX app on `127.0.0.1:8000`), `lmstudio`
+(`:1234`), and `mlx` (`mlx_lm.server` on `:8080`). They are billed as local.
+oMLX requires its API key: `graff key set omlx <key>` or `OMLX_API_KEY`.
+`graff models` lists the models oMLX serves, `graff --model omlx` starts on
+the first one, and `graff --model omlx/<model>` picks one. `/effort` is sent
+as `reasoning_effort`.
+
</details>
## Evaluation results
diff --git a/examples/edge-worker/src/remote.ts b/examples/edge-worker/src/remote.ts
index cbefa804..e82cda2a 100644
--- a/examples/edge-worker/src/remote.ts
+++ b/examples/edge-worker/src/remote.ts
@@ -9,9 +9,9 @@
export const HARNESS_VERSION = "0.13";
-export type ModelName = "MiniMax-M2.5" | "MiniMax-M2.7" | "MiniMax-M3" | "accounts/fireworks/models/deepseek-v4-flash" | "accounts/fireworks/models/deepseek-v4-pro" | "accounts/fireworks/models/glm-5p2" | "accounts/fireworks/models/gpt-oss-120b" | "accounts/fireworks/models/kimi-k2p6" | "accounts/fireworks/models/kimi-k2p7-code" | "accounts/fireworks/models/minimax-m3" | "accounts/fireworks/models/qwen3p7-plus" | "alibaba/qwen3.8-27b" | "anthropic/claude-sonnet-4.6" | "claude-fable-5" | "claude-fable-5-1" | "claude-haiku-5-5" | "claude-opus-4-5" | "claude-opus-4-6" | "claude-opus-4-7" | "claude-opus-4-8" | "claude-opus-4.8" | "claude-opus-5" | "claude-opus-5-5" | "claude-sonnet-4-5" | "claude-sonnet-4-6" | "claude-sonnet-4.6" | "claude-sonnet-5" | "claude-sonnet-5-5" | "deepseek-chat" | "deepseek-reasoner" | "deepseek-v4-flash" | "deepseek-v4-pro" | "fugu" | "fugu-ultra" | "fugu-ultra-20260615" | "gemini-3-flash-preview" | "gemini-3.1-flash-lite" | "gemini-3.1-pro-preview" | "gemini-3.5-flash" | "gemini-3.5-flash-lite" | "gemini-3.6-flash" | "gemini-3.7-flash" | "gemini-3.8-flash" | "gemini-4-argon" | "gemma-4-26b-a4b-it" | "gemma-4-31b" | "gemma-4-31b-it" | "glm-4.5" | "glm-4.7" | "glm-5" | "glm-5-turbo" | "glm-5.2" | "glm-5.3" | "glm-5.3-flash" | "glm-5v-turbo" | "gpt-5-codex" | "gpt-5.2" | "gpt-5.3-codex-spark" | "gpt-5.4" | "gpt-5.4-mini" | "gpt-5.4-pro" | "gpt-5.5" | "gpt-5.6" | "gpt-5.6-luna" | "gpt-5.6-sol" | "gpt-5.6-terra" | "gpt-6-astra" | "gpt-6-luna" | "gpt-6-sol" | "gpt-6.1-sol" | "gpt-oss-120b" | "grok-4.3" | "grok-4.6" | "grok-4.7" | "grok-build" | "k3" | "kilo-auto/small" | "kimi-for-coding" | "kimi-for-coding-highspeed" | "kimi-k2.6" | "kimi-k2.7-code" | "kimi-k3" | "kimi-latest" | "lmstudio" | "mimo-v2-flash" | "mimo-v2.5" | "mimo-v2.5-pro" | "mimo-v2.5-pro-ultraspeed" | "mimo-v2.6-flash" | "mimo-v2.6-pro" | "mimo-v2.6-pro-ultraspeed" | "minimax-m3" | "mistral-medium-latest" | "mlx-community/Qwen3.6-27B-OptiQ-4bit" | "muse-spark-1.2" | "muse-spark-1.2-contributor" | "openai/gpt-oss-120b" | (string & {});
+export type ModelName = "MiniMax-M2.5" | "MiniMax-M2.7" | "MiniMax-M3" | "accounts/fireworks/models/deepseek-v4-flash" | "accounts/fireworks/models/deepseek-v4-pro" | "accounts/fireworks/models/glm-5p2" | "accounts/fireworks/models/gpt-oss-120b" | "accounts/fireworks/models/kimi-k2p6" | "accounts/fireworks/models/kimi-k2p7-code" | "accounts/fireworks/models/minimax-m3" | "accounts/fireworks/models/qwen3p7-plus" | "alibaba/qwen3.8-27b" | "anthropic/claude-sonnet-4.6" | "claude-fable-5" | "claude-fable-5-1" | "claude-haiku-5-5" | "claude-opus-4-5" | "claude-opus-4-6" | "claude-opus-4-7" | "claude-opus-4-8" | "claude-opus-4.8" | "claude-opus-5" | "claude-opus-5-5" | "claude-sonnet-4-5" | "claude-sonnet-4-6" | "claude-sonnet-4.6" | "claude-sonnet-5" | "claude-sonnet-5-5" | "deepseek-chat" | "deepseek-reasoner" | "deepseek-v4-flash" | "deepseek-v4-pro" | "fugu" | "fugu-ultra" | "fugu-ultra-20260615" | "gemini-3-flash-preview" | "gemini-3.1-flash-lite" | "gemini-3.1-pro-preview" | "gemini-3.5-flash" | "gemini-3.5-flash-lite" | "gemini-3.6-flash" | "gemini-3.7-flash" | "gemini-3.8-flash" | "gemini-4-argon" | "gemma-4-26b-a4b-it" | "gemma-4-31b" | "gemma-4-31b-it" | "glm-4.5" | "glm-4.7" | "glm-5" | "glm-5-turbo" | "glm-5.2" | "glm-5.3" | "glm-5.3-flash" | "glm-5v-turbo" | "gpt-5-codex" | "gpt-5.2" | "gpt-5.3-codex-spark" | "gpt-5.4" | "gpt-5.4-mini" | "gpt-5.4-pro" | "gpt-5.5" | "gpt-5.6" | "gpt-5.6-luna" | "gpt-5.6-sol" | "gpt-5.6-terra" | "gpt-6-astra" | "gpt-6-luna" | "gpt-6-sol" | "gpt-6.1-sol" | "gpt-oss-120b" | "grok-4.3" | "grok-4.6" | "grok-4.7" | "grok-build" | "k3" | "kilo-auto/small" | "kimi-for-coding" | "kimi-for-coding-highspeed" | "kimi-k2.6" | "kimi-k2.7-code" | "kimi-k3" | "kimi-latest" | "lmstudio" | "mimo-v2-flash" | "mimo-v2.5" | "mimo-v2.5-pro" | "mimo-v2.5-pro-ultraspeed" | "mimo-v2.6-flash" | "mimo-v2.6-pro" | "mimo-v2.6-pro-ultraspeed" | "minimax-m3" | "mistral-medium-latest" | "mlx-community/Qwen3.6-27B-OptiQ-4bit" | "muse-spark-1.2" | "muse-spark-1.2-contributor" | "omlx" | "openai/gpt-oss-120b" | (string & {});
export type ToolName = "shell" | "read_file" | "edit_file" | "write_file" | "webfetch" | "skill" | "codedb" | "read_tool_result" | "render_html" | "todo_write" | "todo_read" | "eval" | "note_constraint" | "ask_user" | "attempt_completion" | "load_tool_schemas" | "mcp_search_tools" | "mcp_select_tool" | "clock_sleep" | "subagent" | "workflow" | "agent_output" | "agent_message" | "subagent_resume" | "learn_candidate" | "peer_message" | "workspace" | "yield_turn" | "imagegen" | "rlm";
-export type ProviderId = "anthropic" | "codegraff" | "deepseek" | "openai" | "google" | "minimax" | "xiaomi" | "kilo" | "groq" | "cerebras" | "mistral" | "kimi" | "moonshot" | "xai" | "zai" | "meta" | "vercel" | "openrouter" | "fugu" | "fireworks" | "mlx" | "lmstudio" | "chatgpt-new" | "codex";
+export type ProviderId = "anthropic" | "codegraff" | "deepseek" | "openai" | "google" | "minimax" | "xiaomi" | "kilo" | "groq" | "cerebras" | "mistral" | "kimi" | "moonshot" | "xai" | "zai" | "meta" | "vercel" | "openrouter" | "fugu" | "fireworks" | "mlx" | "lmstudio" | "omlx" | "chatgpt-new" | "codex";
/** Events streamed by the bridge (same `--json` contract as the stdio SDK).
* `seq` is a monotonic, gap-free, 1-based per-session id the bridge forwards
@@ -25,7 +25,7 @@ export type Event =
| { seq: number; type: "reasoning"; text: string }
| { seq: number; type: "started"; provider: string; model: string }
| { seq: number; type: "model_call_started"; provider: string; model: string }
- | { seq: number; type: "model_call_finished"; provider: string; model: string; ok: boolean; ms: number }
+ | { seq: number; type: "model_call_finished"; provider: string; model: string; ok: boolean; ms: number; retried?: boolean }
| { seq: number; type: "tool_call"; name: string; input: Record<string, unknown> }
| { seq: number; type: "tool_call_started"; name: string; input: Record<string, unknown> }
| { seq: number; type: "tool_rejected"; name: string; reason: "budget" | "duplicate" | string; input: Record<string, unknown>; message: string }
@@ -528,5 +528,5 @@ export async function* runAgentRemote(opts: RunAgentRemoteOptions): AsyncGenerat
}
}
-export const MODELS: ModelName[] = ["MiniMax-M2.5", "MiniMax-M2.7", "MiniMax-M3", "accounts/fireworks/models/deepseek-v4-flash", "accounts/fireworks/models/deepseek-v4-pro", "accounts/fireworks/models/glm-5p2", "accounts/fireworks/models/gpt-oss-120b", "accounts/fireworks/models/kimi-k2p6", "accounts/fireworks/models/kimi-k2p7-code", "accounts/fireworks/models/minimax-m3", "accounts/fireworks/models/qwen3p7-plus", "alibaba/qwen3.8-27b", "anthropic/claude-sonnet-4.6", "claude-fable-5", "claude-fable-5-1", "claude-haiku-5-5", "claude-opus-4-5", "claude-opus-4-6", "claude-opus-4-7", "claude-opus-4-8", "claude-opus-4.8", "claude-opus-5", "claude-opus-5-5", "claude-sonnet-4-5", "claude-sonnet-4-6", "claude-sonnet-4.6", "claude-sonnet-5", "claude-sonnet-5-5", "deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro", "fugu", "fugu-ultra", "fugu-ultra-20260615", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.6-flash", "gemini-3.7-flash", "gemini-3.8-flash", "gemini-4-argon", "gemma-4-26b-a4b-it", "gemma-4-31b", "gemma-4-31b-it", "glm-4.5", "glm-4.7", "glm-5", "glm-5-turbo", "glm-5.2", "glm-5.3", "glm-5.3-flash", "glm-5v-turbo", "gpt-5-codex", "gpt-5.2", "gpt-5.3-codex-spark", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-pro", "gpt-5.5", "gpt-5.6", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-astra", "gpt-6-luna", "gpt-6-sol", "gpt-6.1-sol", "gpt-oss-120b", "grok-4.3", "grok-4.6", "grok-4.7", "grok-build", "k3", "kilo-auto/small", "kimi-for-coding", "kimi-for-coding-highspeed", "kimi-k2.6", "kimi-k2.7-code", "kimi-k3", "kimi-latest", "lmstudio", "mimo-v2-flash", "mimo-v2.5", "mimo-v2.5-pro", "mimo-v2.5-pro-ultraspeed", "mimo-v2.6-flash", "mimo-v2.6-pro", "mimo-v2.6-pro-ultraspeed", "minimax-m3", "mistral-medium-latest", "mlx-community/Qwen3.6-27B-OptiQ-4bit", "muse-spark-1.2", "muse-spark-1.2-contributor", "openai/gpt-oss-120b"];
+export const MODELS: ModelName[] = ["MiniMax-M2.5", "MiniMax-M2.7", "MiniMax-M3", "accounts/fireworks/models/deepseek-v4-flash", "accounts/fireworks/models/deepseek-v4-pro", "accounts/fireworks/models/glm-5p2", "accounts/fireworks/models/gpt-oss-120b", "accounts/fireworks/models/kimi-k2p6", "accounts/fireworks/models/kimi-k2p7-code", "accounts/fireworks/models/minimax-m3", "accounts/fireworks/models/qwen3p7-plus", "alibaba/qwen3.8-27b", "anthropic/claude-sonnet-4.6", "claude-fable-5", "claude-fable-5-1", "claude-haiku-5-5", "claude-opus-4-5", "claude-opus-4-6", "claude-opus-4-7", "claude-opus-4-8", "claude-opus-4.8", "claude-opus-5", "claude-opus-5-5", "claude-sonnet-4-5", "claude-sonnet-4-6", "claude-sonnet-4.6", "claude-sonnet-5", "claude-sonnet-5-5", "deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro", "fugu", "fugu-ultra", "fugu-ultra-20260615", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.6-flash", "gemini-3.7-flash", "gemini-3.8-flash", "gemini-4-argon", "gemma-4-26b-a4b-it", "gemma-4-31b", "gemma-4-31b-it", "glm-4.5", "glm-4.7", "glm-5", "glm-5-turbo", "glm-5.2", "glm-5.3", "glm-5.3-flash", "glm-5v-turbo", "gpt-5-codex", "gpt-5.2", "gpt-5.3-codex-spark", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-pro", "gpt-5.5", "gpt-5.6", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-astra", "gpt-6-luna", "gpt-6-sol", "gpt-6.1-sol", "gpt-oss-120b", "grok-4.3", "grok-4.6", "grok-4.7", "grok-build", "k3", "kilo-auto/small", "kimi-for-coding", "kimi-for-coding-highspeed", "kimi-k2.6", "kimi-k2.7-code", "kimi-k3", "kimi-latest", "lmstudio", "mimo-v2-flash", "mimo-v2.5", "mimo-v2.5-pro", "mimo-v2.5-pro-ultraspeed", "mimo-v2.6-flash", "mimo-v2.6-pro", "mimo-v2.6-pro-ultraspeed", "minimax-m3", "mistral-medium-latest", "mlx-community/Qwen3.6-27B-OptiQ-4bit", "muse-spark-1.2", "muse-spark-1.2-contributor", "omlx", "openai/gpt-oss-120b"];
export const TOOLS: ToolName[] = ["shell", "read_file", "edit_file", "write_file", "webfetch", "skill", "codedb", "read_tool_result", "render_html", "todo_write", "todo_read", "eval", "note_constraint", "ask_user", "attempt_completion", "load_tool_schemas", "mcp_search_tools", "mcp_select_tool", "clock_sleep", "subagent", "workflow", "agent_output", "agent_message", "subagent_resume", "learn_candidate", "peer_message", "workspace", "yield_turn", "imagegen", "rlm"];
diff --git a/lean-proofs/Graff/Provider.lean b/lean-proofs/Graff/Provider.lean
index 46a74f27..0bb93be9 100644
--- a/lean-proofs/Graff/Provider.lean
+++ b/lean-proofs/Graff/Provider.lean
@@ -1,7 +1,7 @@
/-
Provider table kernel.
- Twenty-four baked rows. Each row is data: an id plus the axes `Kind`,
+ Twenty-five baked rows. Each row is data: an id plus the axes `Kind`,
`Auth`, login, catalog, and two bools. The program does not branch per
brand. Adding a vendor is one constructor here and one line in
`provider.zig`; the Zig fixture test is the ratchet.
@@ -69,6 +69,7 @@ def specs : List Row :=
, { id := "fireworks", kind := .openai, auth := .bearer, catalog := .openai }
, { id := "mlx", kind := .openai, auth := .bearer }
, { id := "lmstudio", kind := .openai, auth := .bearer }
+ , { id := "omlx", kind := .openai, auth := .bearer, catalog := .openai, takesEffort := true }
, { id := "chatgpt-new", kind := .responses, auth := .bearer, login := .chatgptBrowser, catalog := .openai, subLogin := true }
, { id := "codex", kind := .responses, auth := .bearer, login := .codexDevice, catalog := .codex, subLogin := true }
]
@@ -89,7 +90,7 @@ def xApiKeyCount : Nat := countWhere (fun r => r.auth == .xApiKey) specs
def providerRows : Nat := specs.length
-theorem provider_cube : providerRows = 24 := by native_decide
+theorem provider_cube : providerRows = 25 := by native_decide
theorem responses_vendors : responsesCount = 3 := by native_decide
theorem one_x_api_key : xApiKeyCount = 1 := by native_decide
diff --git a/lean-proofs/KERNELS.md b/lean-proofs/KERNELS.md
index e21dc96c..314ecb8e 100644
--- a/lean-proofs/KERNELS.md
+++ b/lean-proofs/KERNELS.md
@@ -46,7 +46,7 @@ not a skip: TUI, prompts, and SSE bytes stay out of Lean on purpose.
|---|---|---|
| `ToolCatalog` | 64 flag cubes | MCP names, descriptions, JSON wrappers |
| `Transport` | 96 turns + `Event`/`step`; **1** WS cell | frames, TLS, idle-kill timing of the server |
-| `Provider` | 21 baked rows | live `/models` overlay, Kimi protocol flip |
+| `Provider` | 25 baked rows | live `/models` overlay, Kimi protocol flip |
| `GoalLoop` | 360 gate cells + `Event`/`step` | model wording, whether the work is correct |
| `PromptCache` | 48 cells + `Event`/`step` | provider cache HIT, uuid5 cwd bytes, vision pin |
| `PromptPrefix` | 6 cells + `Event`/`step` | catalog wording, provider HIT, compaction |
diff --git a/sdk/py/harness_sdk.py b/sdk/py/harness_sdk.py
index d48a4650..6c7bee81 100644
--- a/sdk/py/harness_sdk.py
+++ b/sdk/py/harness_sdk.py
@@ -13,9 +13,9 @@ import urllib.error
import urllib.request
from typing import Iterator, List, Literal, Optional, TypedDict, Union
-MODELS = ["MiniMax-M2.5", "MiniMax-M2.7", "MiniMax-M3", "accounts/fireworks/models/deepseek-v4-flash", "accounts/fireworks/models/deepseek-v4-pro", "accounts/fireworks/models/glm-5p2", "accounts/fireworks/models/gpt-oss-120b", "accounts/fireworks/models/kimi-k2p6", "accounts/fireworks/models/kimi-k2p7-code", "accounts/fireworks/models/minimax-m3", "accounts/fireworks/models/qwen3p7-plus", "alibaba/qwen3.8-27b", "anthropic/claude-sonnet-4.6", "claude-fable-5", "claude-fable-5-1", "claude-haiku-5-5", "claude-opus-4-5", "claude-opus-4-6", "claude-opus-4-7", "claude-opus-4-8", "claude-opus-4.8", "claude-opus-5", "claude-opus-5-5", "claude-sonnet-4-5", "claude-sonnet-4-6", "claude-sonnet-4.6", "claude-sonnet-5", "claude-sonnet-5-5", "deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro", "fugu", "fugu-ultra", "fugu-ultra-20260615", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.6-flash", "gemini-3.7-flash", "gemini-3.8-flash", "gemini-4-argon", "gemma-4-26b-a4b-it", "gemma-4-31b", "gemma-4-31b-it", "glm-4.5", "glm-4.7", "glm-5", "glm-5-turbo", "glm-5.2", "glm-5.3", "glm-5.3-flash", "glm-5v-turbo", "gpt-5-codex", "gpt-5.2", "gpt-5.3-codex-spark", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-pro", "gpt-5.5", "gpt-5.6", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-astra", "gpt-6-luna", "gpt-6-sol", "gpt-6.1-sol", "gpt-oss-120b", "grok-4.3", "grok-4.6", "grok-4.7", "grok-build", "k3", "kilo-auto/small", "kimi-for-coding", "kimi-for-coding-highspeed", "kimi-k2.6", "kimi-k2.7-code", "kimi-k3", "kimi-latest", "lmstudio", "mimo-v2-flash", "mimo-v2.5", "mimo-v2.5-pro", "mimo-v2.5-pro-ultraspeed", "mimo-v2.6-flash", "mimo-v2.6-pro", "mimo-v2.6-pro-ultraspeed", "minimax-m3", "mistral-medium-latest", "mlx-community/Qwen3.6-27B-OptiQ-4bit", "muse-spark-1.2", "muse-spark-1.2-contributor", "openai/gpt-oss-120b"]
+MODELS = ["MiniMax-M2.5", "MiniMax-M2.7", "MiniMax-M3", "accounts/fireworks/models/deepseek-v4-flash", "accounts/fireworks/models/deepseek-v4-pro", "accounts/fireworks/models/glm-5p2", "accounts/fireworks/models/gpt-oss-120b", "accounts/fireworks/models/kimi-k2p6", "accounts/fireworks/models/kimi-k2p7-code", "accounts/fireworks/models/minimax-m3", "accounts/fireworks/models/qwen3p7-plus", "alibaba/qwen3.8-27b", "anthropic/claude-sonnet-4.6", "claude-fable-5", "claude-fable-5-1", "claude-haiku-5-5", "claude-opus-4-5", "claude-opus-4-6", "claude-opus-4-7", "claude-opus-4-8", "claude-opus-4.8", "claude-opus-5", "claude-opus-5-5", "claude-sonnet-4-5", "claude-sonnet-4-6", "claude-sonnet-4.6", "claude-sonnet-5", "claude-sonnet-5-5", "deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro", "fugu", "fugu-ultra", "fugu-ultra-20260615", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.6-flash", "gemini-3.7-flash", "gemini-3.8-flash", "gemini-4-argon", "gemma-4-26b-a4b-it", "gemma-4-31b", "gemma-4-31b-it", "glm-4.5", "glm-4.7", "glm-5", "glm-5-turbo", "glm-5.2", "glm-5.3", "glm-5.3-flash", "glm-5v-turbo", "gpt-5-codex", "gpt-5.2", "gpt-5.3-codex-spark", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-pro", "gpt-5.5", "gpt-5.6", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-astra", "gpt-6-luna", "gpt-6-sol", "gpt-6.1-sol", "gpt-oss-120b", "grok-4.3", "grok-4.6", "grok-4.7", "grok-build", "k3", "kilo-auto/small", "kimi-for-coding", "kimi-for-coding-highspeed", "kimi-k2.6", "kimi-k2.7-code", "kimi-k3", "kimi-latest", "lmstudio", "mimo-v2-flash", "mimo-v2.5", "mimo-v2.5-pro", "mimo-v2.5-pro-ultraspeed", "mimo-v2.6-flash", "mimo-v2.6-pro", "mimo-v2.6-pro-ultraspeed", "minimax-m3", "mistral-medium-latest", "mlx-community/Qwen3.6-27B-OptiQ-4bit", "muse-spark-1.2", "muse-spark-1.2-contributor", "omlx", "openai/gpt-oss-120b"]
TOOLS = ["shell", "read_file", "edit_file", "write_file", "webfetch", "skill", "codedb", "read_tool_result", "render_html", "todo_write", "todo_read", "eval", "note_constraint", "ask_user", "attempt_completion", "load_tool_schemas", "mcp_search_tools", "mcp_select_tool", "clock_sleep", "subagent", "workflow", "agent_output", "agent_message", "subagent_resume", "learn_candidate", "peer_message", "workspace", "yield_turn", "imagegen", "rlm"]
-PROVIDERS = ["anthropic", "codegraff", "deepseek", "openai", "google", "minimax", "xiaomi", "kilo", "groq", "cerebras", "mistral", "kimi", "moonshot", "xai", "zai", "meta", "vercel", "openrouter", "fugu", "fireworks", "mlx", "lmstudio", "chatgpt-new", "codex"]
+PROVIDERS = ["anthropic", "codegraff", "deepseek", "openai", "google", "minimax", "xiaomi", "kilo", "groq", "cerebras", "mistral", "kimi", "moonshot", "xai", "zai", "meta", "vercel", "openrouter", "fugu", "fireworks", "mlx", "lmstudio", "omlx", "chatgpt-new", "codex"]
class ImageUrlInput(TypedDict):
diff --git a/sdk/ts/harness.ts b/sdk/ts/harness.ts
index fafddb13..02068c96 100644
--- a/sdk/ts/harness.ts
+++ b/sdk/ts/harness.ts
@@ -11,9 +11,9 @@ import type { Readable, Writable } from "node:stream";
export const HARNESS_VERSION = "0.13";
-export type ModelName = "MiniMax-M2.5" | "MiniMax-M2.7" | "MiniMax-M3" | "accounts/fireworks/models/deepseek-v4-flash" | "accounts/fireworks/models/deepseek-v4-pro" | "accounts/fireworks/models/glm-5p2" | "accounts/fireworks/models/gpt-oss-120b" | "accounts/fireworks/models/kimi-k2p6" | "accounts/fireworks/models/kimi-k2p7-code" | "accounts/fireworks/models/minimax-m3" | "accounts/fireworks/models/qwen3p7-plus" | "alibaba/qwen3.8-27b" | "anthropic/claude-sonnet-4.6" | "claude-fable-5" | "claude-fable-5-1" | "claude-haiku-5-5" | "claude-opus-4-5" | "claude-opus-4-6" | "claude-opus-4-7" | "claude-opus-4-8" | "claude-opus-4.8" | "claude-opus-5" | "claude-opus-5-5" | "claude-sonnet-4-5" | "claude-sonnet-4-6" | "claude-sonnet-4.6" | "claude-sonnet-5" | "claude-sonnet-5-5" | "deepseek-chat" | "deepseek-reasoner" | "deepseek-v4-flash" | "deepseek-v4-pro" | "fugu" | "fugu-ultra" | "fugu-ultra-20260615" | "gemini-3-flash-preview" | "gemini-3.1-flash-lite" | "gemini-3.1-pro-preview" | "gemini-3.5-flash" | "gemini-3.5-flash-lite" | "gemini-3.6-flash" | "gemini-3.7-flash" | "gemini-3.8-flash" | "gemini-4-argon" | "gemma-4-26b-a4b-it" | "gemma-4-31b" | "gemma-4-31b-it" | "glm-4.5" | "glm-4.7" | "glm-5" | "glm-5-turbo" | "glm-5.2" | "glm-5.3" | "glm-5.3-flash" | "glm-5v-turbo" | "gpt-5-codex" | "gpt-5.2" | "gpt-5.3-codex-spark" | "gpt-5.4" | "gpt-5.4-mini" | "gpt-5.4-pro" | "gpt-5.5" | "gpt-5.6" | "gpt-5.6-luna" | "gpt-5.6-sol" | "gpt-5.6-terra" | "gpt-6-astra" | "gpt-6-luna" | "gpt-6-sol" | "gpt-6.1-sol" | "gpt-oss-120b" | "grok-4.3" | "grok-4.6" | "grok-4.7" | "grok-build" | "k3" | "kilo-auto/small" | "kimi-for-coding" | "kimi-for-coding-highspeed" | "kimi-k2.6" | "kimi-k2.7-code" | "kimi-k3" | "kimi-latest" | "lmstudio" | "mimo-v2-flash" | "mimo-v2.5" | "mimo-v2.5-pro" | "mimo-v2.5-pro-ultraspeed" | "mimo-v2.6-flash" | "mimo-v2.6-pro" | "mimo-v2.6-pro-ultraspeed" | "minimax-m3" | "mistral-medium-latest" | "mlx-community/Qwen3.6-27B-OptiQ-4bit" | "muse-spark-1.2" | "muse-spark-1.2-contributor" | "openai/gpt-oss-120b" | (string & {});
+export type ModelName = "MiniMax-M2.5" | "MiniMax-M2.7" | "MiniMax-M3" | "accounts/fireworks/models/deepseek-v4-flash" | "accounts/fireworks/models/deepseek-v4-pro" | "accounts/fireworks/models/glm-5p2" | "accounts/fireworks/models/gpt-oss-120b" | "accounts/fireworks/models/kimi-k2p6" | "accounts/fireworks/models/kimi-k2p7-code" | "accounts/fireworks/models/minimax-m3" | "accounts/fireworks/models/qwen3p7-plus" | "alibaba/qwen3.8-27b" | "anthropic/claude-sonnet-4.6" | "claude-fable-5" | "claude-fable-5-1" | "claude-haiku-5-5" | "claude-opus-4-5" | "claude-opus-4-6" | "claude-opus-4-7" | "claude-opus-4-8" | "claude-opus-4.8" | "claude-opus-5" | "claude-opus-5-5" | "claude-sonnet-4-5" | "claude-sonnet-4-6" | "claude-sonnet-4.6" | "claude-sonnet-5" | "claude-sonnet-5-5" | "deepseek-chat" | "deepseek-reasoner" | "deepseek-v4-flash" | "deepseek-v4-pro" | "fugu" | "fugu-ultra" | "fugu-ultra-20260615" | "gemini-3-flash-preview" | "gemini-3.1-flash-lite" | "gemini-3.1-pro-preview" | "gemini-3.5-flash" | "gemini-3.5-flash-lite" | "gemini-3.6-flash" | "gemini-3.7-flash" | "gemini-3.8-flash" | "gemini-4-argon" | "gemma-4-26b-a4b-it" | "gemma-4-31b" | "gemma-4-31b-it" | "glm-4.5" | "glm-4.7" | "glm-5" | "glm-5-turbo" | "glm-5.2" | "glm-5.3" | "glm-5.3-flash" | "glm-5v-turbo" | "gpt-5-codex" | "gpt-5.2" | "gpt-5.3-codex-spark" | "gpt-5.4" | "gpt-5.4-mini" | "gpt-5.4-pro" | "gpt-5.5" | "gpt-5.6" | "gpt-5.6-luna" | "gpt-5.6-sol" | "gpt-5.6-terra" | "gpt-6-astra" | "gpt-6-luna" | "gpt-6-sol" | "gpt-6.1-sol" | "gpt-oss-120b" | "grok-4.3" | "grok-4.6" | "grok-4.7" | "grok-build" | "k3" | "kilo-auto/small" | "kimi-for-coding" | "kimi-for-coding-highspeed" | "kimi-k2.6" | "kimi-k2.7-code" | "kimi-k3" | "kimi-latest" | "lmstudio" | "mimo-v2-flash" | "mimo-v2.5" | "mimo-v2.5-pro" | "mimo-v2.5-pro-ultraspeed" | "mimo-v2.6-flash" | "mimo-v2.6-pro" | "mimo-v2.6-pro-ultraspeed" | "minimax-m3" | "mistral-medium-latest" | "mlx-community/Qwen3.6-27B-OptiQ-4bit" | "muse-spark-1.2" | "muse-spark-1.2-contributor" | "omlx" | "openai/gpt-oss-120b" | (string & {});
export type ToolName = "shell" | "read_file" | "edit_file" | "write_file" | "webfetch" | "skill" | "codedb" | "read_tool_result" | "render_html" | "todo_write" | "todo_read" | "eval" | "note_constraint" | "ask_user" | "attempt_completion" | "load_tool_schemas" | "mcp_search_tools" | "mcp_select_tool" | "clock_sleep" | "subagent" | "workflow" | "agent_output" | "agent_message" | "subagent_resume" | "learn_candidate" | "peer_message" | "workspace" | "yield_turn" | "imagegen" | "rlm";
-export type ProviderId = "anthropic" | "codegraff" | "deepseek" | "openai" | "google" | "minimax" | "xiaomi" | "kilo" | "groq" | "cerebras" | "mistral" | "kimi" | "moonshot" | "xai" | "zai" | "meta" | "vercel" | "openrouter" | "fugu" | "fireworks" | "mlx" | "lmstudio" | "chatgpt-new" | "codex";
+export type ProviderId = "anthropic" | "codegraff" | "deepseek" | "openai" | "google" | "minimax" | "xiaomi" | "kilo" | "groq" | "cerebras" | "mistral" | "kimi" | "moonshot" | "xai" | "zai" | "meta" | "vercel" | "openrouter" | "fugu" | "fireworks" | "mlx" | "lmstudio" | "omlx" | "chatgpt-new" | "codex";
/** Events streamed by `harness --json` (one JSON object per stdout line), each
* discriminated by `type` — the analogue of codegraff's AgentEvent stream.
@@ -684,5 +684,5 @@ export async function* runAgent(opts: RunAgentOptions): AsyncGenerator<Event> {
}
}
-export const MODELS: ModelName[] = ["MiniMax-M2.5", "MiniMax-M2.7", "MiniMax-M3", "accounts/fireworks/models/deepseek-v4-flash", "accounts/fireworks/models/deepseek-v4-pro", "accounts/fireworks/models/glm-5p2", "accounts/fireworks/models/gpt-oss-120b", "accounts/fireworks/models/kimi-k2p6", "accounts/fireworks/models/kimi-k2p7-code", "accounts/fireworks/models/minimax-m3", "accounts/fireworks/models/qwen3p7-plus", "alibaba/qwen3.8-27b", "anthropic/claude-sonnet-4.6", "claude-fable-5", "claude-fable-5-1", "claude-haiku-5-5", "claude-opus-4-5", "claude-opus-4-6", "claude-opus-4-7", "claude-opus-4-8", "claude-opus-4.8", "claude-opus-5", "claude-opus-5-5", "claude-sonnet-4-5", "claude-sonnet-4-6", "claude-sonnet-4.6", "claude-sonnet-5", "claude-sonnet-5-5", "deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro", "fugu", "fugu-ultra", "fugu-ultra-20260615", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.6-flash", "gemini-3.7-flash", "gemini-3.8-flash", "gemini-4-argon", "gemma-4-26b-a4b-it", "gemma-4-31b", "gemma-4-31b-it", "glm-4.5", "glm-4.7", "glm-5", "glm-5-turbo", "glm-5.2", "glm-5.3", "glm-5.3-flash", "glm-5v-turbo", "gpt-5-codex", "gpt-5.2", "gpt-5.3-codex-spark", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-pro", "gpt-5.5", "gpt-5.6", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-astra", "gpt-6-luna", "gpt-6-sol", "gpt-6.1-sol", "gpt-oss-120b", "grok-4.3", "grok-4.6", "grok-4.7", "grok-build", "k3", "kilo-auto/small", "kimi-for-coding", "kimi-for-coding-highspeed", "kimi-k2.6", "kimi-k2.7-code", "kimi-k3", "kimi-latest", "lmstudio", "mimo-v2-flash", "mimo-v2.5", "mimo-v2.5-pro", "mimo-v2.5-pro-ultraspeed", "mimo-v2.6-flash", "mimo-v2.6-pro", "mimo-v2.6-pro-ultraspeed", "minimax-m3", "mistral-medium-latest", "mlx-community/Qwen3.6-27B-OptiQ-4bit", "muse-spark-1.2", "muse-spark-1.2-contributor", "openai/gpt-oss-120b"];
+export const MODELS: ModelName[] = ["MiniMax-M2.5", "MiniMax-M2.7", "MiniMax-M3", "accounts/fireworks/models/deepseek-v4-flash", "accounts/fireworks/models/deepseek-v4-pro", "accounts/fireworks/models/glm-5p2", "accounts/fireworks/models/gpt-oss-120b", "accounts/fireworks/models/kimi-k2p6", "accounts/fireworks/models/kimi-k2p7-code", "accounts/fireworks/models/minimax-m3", "accounts/fireworks/models/qwen3p7-plus", "alibaba/qwen3.8-27b", "anthropic/claude-sonnet-4.6", "claude-fable-5", "claude-fable-5-1", "claude-haiku-5-5", "claude-opus-4-5", "claude-opus-4-6", "claude-opus-4-7", "claude-opus-4-8", "claude-opus-4.8", "claude-opus-5", "claude-opus-5-5", "claude-sonnet-4-5", "claude-sonnet-4-6", "claude-sonnet-4.6", "claude-sonnet-5", "claude-sonnet-5-5", "deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro", "fugu", "fugu-ultra", "fugu-ultra-20260615", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.6-flash", "gemini-3.7-flash", "gemini-3.8-flash", "gemini-4-argon", "gemma-4-26b-a4b-it", "gemma-4-31b", "gemma-4-31b-it", "glm-4.5", "glm-4.7", "glm-5", "glm-5-turbo", "glm-5.2", "glm-5.3", "glm-5.3-flash", "glm-5v-turbo", "gpt-5-codex", "gpt-5.2", "gpt-5.3-codex-spark", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-pro", "gpt-5.5", "gpt-5.6", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-astra", "gpt-6-luna", "gpt-6-sol", "gpt-6.1-sol", "gpt-oss-120b", "grok-4.3", "grok-4.6", "grok-4.7", "grok-build", "k3", "kilo-auto/small", "kimi-for-coding", "kimi-for-coding-highspeed", "kimi-k2.6", "kimi-k2.7-code", "kimi-k3", "kimi-latest", "lmstudio", "mimo-v2-flash", "mimo-v2.5", "mimo-v2.5-pro", "mimo-v2.5-pro-ultraspeed", "mimo-v2.6-flash", "mimo-v2.6-pro", "mimo-v2.6-pro-ultraspeed", "minimax-m3", "mistral-medium-latest", "mlx-community/Qwen3.6-27B-OptiQ-4bit", "muse-spark-1.2", "muse-spark-1.2-contributor", "omlx", "openai/gpt-oss-120b"];
export const TOOLS: ToolName[] = ["shell", "read_file", "edit_file", "write_file", "webfetch", "skill", "codedb", "read_tool_result", "render_html", "todo_write", "todo_read", "eval", "note_constraint", "ask_user", "attempt_completion", "load_tool_schemas", "mcp_search_tools", "mcp_select_tool", "clock_sleep", "subagent", "workflow", "agent_output", "agent_message", "subagent_resume", "learn_candidate", "peer_message", "workspace", "yield_turn", "imagegen", "rlm"];
diff --git a/sdk/ts/remote.ts b/sdk/ts/remote.ts
index d2e85b6c..e82cda2a 100644
--- a/sdk/ts/remote.ts
+++ b/sdk/ts/remote.ts
@@ -9,9 +9,9 @@
export const HARNESS_VERSION = "0.13";
-export type ModelName = "MiniMax-M2.5" | "MiniMax-M2.7" | "MiniMax-M3" | "accounts/fireworks/models/deepseek-v4-flash" | "accounts/fireworks/models/deepseek-v4-pro" | "accounts/fireworks/models/glm-5p2" | "accounts/fireworks/models/gpt-oss-120b" | "accounts/fireworks/models/kimi-k2p6" | "accounts/fireworks/models/kimi-k2p7-code" | "accounts/fireworks/models/minimax-m3" | "accounts/fireworks/models/qwen3p7-plus" | "alibaba/qwen3.8-27b" | "anthropic/claude-sonnet-4.6" | "claude-fable-5" | "claude-fable-5-1" | "claude-haiku-5-5" | "claude-opus-4-5" | "claude-opus-4-6" | "claude-opus-4-7" | "claude-opus-4-8" | "claude-opus-4.8" | "claude-opus-5" | "claude-opus-5-5" | "claude-sonnet-4-5" | "claude-sonnet-4-6" | "claude-sonnet-4.6" | "claude-sonnet-5" | "claude-sonnet-5-5" | "deepseek-chat" | "deepseek-reasoner" | "deepseek-v4-flash" | "deepseek-v4-pro" | "fugu" | "fugu-ultra" | "fugu-ultra-20260615" | "gemini-3-flash-preview" | "gemini-3.1-flash-lite" | "gemini-3.1-pro-preview" | "gemini-3.5-flash" | "gemini-3.5-flash-lite" | "gemini-3.6-flash" | "gemini-3.7-flash" | "gemini-3.8-flash" | "gemini-4-argon" | "gemma-4-26b-a4b-it" | "gemma-4-31b" | "gemma-4-31b-it" | "glm-4.5" | "glm-4.7" | "glm-5" | "glm-5-turbo" | "glm-5.2" | "glm-5.3" | "glm-5.3-flash" | "glm-5v-turbo" | "gpt-5-codex" | "gpt-5.2" | "gpt-5.3-codex-spark" | "gpt-5.4" | "gpt-5.4-mini" | "gpt-5.4-pro" | "gpt-5.5" | "gpt-5.6" | "gpt-5.6-luna" | "gpt-5.6-sol" | "gpt-5.6-terra" | "gpt-6-astra" | "gpt-6-luna" | "gpt-6-sol" | "gpt-6.1-sol" | "gpt-oss-120b" | "grok-4.3" | "grok-4.6" | "grok-4.7" | "grok-build" | "k3" | "kilo-auto/small" | "kimi-for-coding" | "kimi-for-coding-highspeed" | "kimi-k2.6" | "kimi-k2.7-code" | "kimi-k3" | "kimi-latest" | "lmstudio" | "mimo-v2-flash" | "mimo-v2.5" | "mimo-v2.5-pro" | "mimo-v2.5-pro-ultraspeed" | "mimo-v2.6-flash" | "mimo-v2.6-pro" | "mimo-v2.6-pro-ultraspeed" | "minimax-m3" | "mistral-medium-latest" | "mlx-community/Qwen3.6-27B-OptiQ-4bit" | "muse-spark-1.2" | "muse-spark-1.2-contributor" | "openai/gpt-oss-120b" | (string & {});
+export type ModelName = "MiniMax-M2.5" | "MiniMax-M2.7" | "MiniMax-M3" | "accounts/fireworks/models/deepseek-v4-flash" | "accounts/fireworks/models/deepseek-v4-pro" | "accounts/fireworks/models/glm-5p2" | "accounts/fireworks/models/gpt-oss-120b" | "accounts/fireworks/models/kimi-k2p6" | "accounts/fireworks/models/kimi-k2p7-code" | "accounts/fireworks/models/minimax-m3" | "accounts/fireworks/models/qwen3p7-plus" | "alibaba/qwen3.8-27b" | "anthropic/claude-sonnet-4.6" | "claude-fable-5" | "claude-fable-5-1" | "claude-haiku-5-5" | "claude-opus-4-5" | "claude-opus-4-6" | "claude-opus-4-7" | "claude-opus-4-8" | "claude-opus-4.8" | "claude-opus-5" | "claude-opus-5-5" | "claude-sonnet-4-5" | "claude-sonnet-4-6" | "claude-sonnet-4.6" | "claude-sonnet-5" | "claude-sonnet-5-5" | "deepseek-chat" | "deepseek-reasoner" | "deepseek-v4-flash" | "deepseek-v4-pro" | "fugu" | "fugu-ultra" | "fugu-ultra-20260615" | "gemini-3-flash-preview" | "gemini-3.1-flash-lite" | "gemini-3.1-pro-preview" | "gemini-3.5-flash" | "gemini-3.5-flash-lite" | "gemini-3.6-flash" | "gemini-3.7-flash" | "gemini-3.8-flash" | "gemini-4-argon" | "gemma-4-26b-a4b-it" | "gemma-4-31b" | "gemma-4-31b-it" | "glm-4.5" | "glm-4.7" | "glm-5" | "glm-5-turbo" | "glm-5.2" | "glm-5.3" | "glm-5.3-flash" | "glm-5v-turbo" | "gpt-5-codex" | "gpt-5.2" | "gpt-5.3-codex-spark" | "gpt-5.4" | "gpt-5.4-mini" | "gpt-5.4-pro" | "gpt-5.5" | "gpt-5.6" | "gpt-5.6-luna" | "gpt-5.6-sol" | "gpt-5.6-terra" | "gpt-6-astra" | "gpt-6-luna" | "gpt-6-sol" | "gpt-6.1-sol" | "gpt-oss-120b" | "grok-4.3" | "grok-4.6" | "grok-4.7" | "grok-build" | "k3" | "kilo-auto/small" | "kimi-for-coding" | "kimi-for-coding-highspeed" | "kimi-k2.6" | "kimi-k2.7-code" | "kimi-k3" | "kimi-latest" | "lmstudio" | "mimo-v2-flash" | "mimo-v2.5" | "mimo-v2.5-pro" | "mimo-v2.5-pro-ultraspeed" | "mimo-v2.6-flash" | "mimo-v2.6-pro" | "mimo-v2.6-pro-ultraspeed" | "minimax-m3" | "mistral-medium-latest" | "mlx-community/Qwen3.6-27B-OptiQ-4bit" | "muse-spark-1.2" | "muse-spark-1.2-contributor" | "omlx" | "openai/gpt-oss-120b" | (string & {});
export type ToolName = "shell" | "read_file" | "edit_file" | "write_file" | "webfetch" | "skill" | "codedb" | "read_tool_result" | "render_html" | "todo_write" | "todo_read" | "eval" | "note_constraint" | "ask_user" | "attempt_completion" | "load_tool_schemas" | "mcp_search_tools" | "mcp_select_tool" | "clock_sleep" | "subagent" | "workflow" | "agent_output" | "agent_message" | "subagent_resume" | "learn_candidate" | "peer_message" | "workspace" | "yield_turn" | "imagegen" | "rlm";
-export type ProviderId = "anthropic" | "codegraff" | "deepseek" | "openai" | "google" | "minimax" | "xiaomi" | "kilo" | "groq" | "cerebras" | "mistral" | "kimi" | "moonshot" | "xai" | "zai" | "meta" | "vercel" | "openrouter" | "fugu" | "fireworks" | "mlx" | "lmstudio" | "chatgpt-new" | "codex";
+export type ProviderId = "anthropic" | "codegraff" | "deepseek" | "openai" | "google" | "minimax" | "xiaomi" | "kilo" | "groq" | "cerebras" | "mistral" | "kimi" | "moonshot" | "xai" | "zai" | "meta" | "vercel" | "openrouter" | "fugu" | "fireworks" | "mlx" | "lmstudio" | "omlx" | "chatgpt-new" | "codex";
/** Events streamed by the bridge (same `--json` contract as the stdio SDK).
* `seq` is a monotonic, gap-free, 1-based per-session id the bridge forwards
@@ -528,5 +528,5 @@ export async function* runAgentRemote(opts: RunAgentRemoteOptions): AsyncGenerat
}
}
-export const MODELS: ModelName[] = ["MiniMax-M2.5", "MiniMax-M2.7", "MiniMax-M3", "accounts/fireworks/models/deepseek-v4-flash", "accounts/fireworks/models/deepseek-v4-pro", "accounts/fireworks/models/glm-5p2", "accounts/fireworks/models/gpt-oss-120b", "accounts/fireworks/models/kimi-k2p6", "accounts/fireworks/models/kimi-k2p7-code", "accounts/fireworks/models/minimax-m3", "accounts/fireworks/models/qwen3p7-plus", "alibaba/qwen3.8-27b", "anthropic/claude-sonnet-4.6", "claude-fable-5", "claude-fable-5-1", "claude-haiku-5-5", "claude-opus-4-5", "claude-opus-4-6", "claude-opus-4-7", "claude-opus-4-8", "claude-opus-4.8", "claude-opus-5", "claude-opus-5-5", "claude-sonnet-4-5", "claude-sonnet-4-6", "claude-sonnet-4.6", "claude-sonnet-5", "claude-sonnet-5-5", "deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro", "fugu", "fugu-ultra", "fugu-ultra-20260615", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.6-flash", "gemini-3.7-flash", "gemini-3.8-flash", "gemini-4-argon", "gemma-4-26b-a4b-it", "gemma-4-31b", "gemma-4-31b-it", "glm-4.5", "glm-4.7", "glm-5", "glm-5-turbo", "glm-5.2", "glm-5.3", "glm-5.3-flash", "glm-5v-turbo", "gpt-5-codex", "gpt-5.2", "gpt-5.3-codex-spark", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-pro", "gpt-5.5", "gpt-5.6", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-astra", "gpt-6-luna", "gpt-6-sol", "gpt-6.1-sol", "gpt-oss-120b", "grok-4.3", "grok-4.6", "grok-4.7", "grok-build", "k3", "kilo-auto/small", "kimi-for-coding", "kimi-for-coding-highspeed", "kimi-k2.6", "kimi-k2.7-code", "kimi-k3", "kimi-latest", "lmstudio", "mimo-v2-flash", "mimo-v2.5", "mimo-v2.5-pro", "mimo-v2.5-pro-ultraspeed", "mimo-v2.6-flash", "mimo-v2.6-pro", "mimo-v2.6-pro-ultraspeed", "minimax-m3", "mistral-medium-latest", "mlx-community/Qwen3.6-27B-OptiQ-4bit", "muse-spark-1.2", "muse-spark-1.2-contributor", "openai/gpt-oss-120b"];
+export const MODELS: ModelName[] = ["MiniMax-M2.5", "MiniMax-M2.7", "MiniMax-M3", "accounts/fireworks/models/deepseek-v4-flash", "accounts/fireworks/models/deepseek-v4-pro", "accounts/fireworks/models/glm-5p2", "accounts/fireworks/models/gpt-oss-120b", "accounts/fireworks/models/kimi-k2p6", "accounts/fireworks/models/kimi-k2p7-code", "accounts/fireworks/models/minimax-m3", "accounts/fireworks/models/qwen3p7-plus", "alibaba/qwen3.8-27b", "anthropic/claude-sonnet-4.6", "claude-fable-5", "claude-fable-5-1", "claude-haiku-5-5", "claude-opus-4-5", "claude-opus-4-6", "claude-opus-4-7", "claude-opus-4-8", "claude-opus-4.8", "claude-opus-5", "claude-opus-5-5", "claude-sonnet-4-5", "claude-sonnet-4-6", "claude-sonnet-4.6", "claude-sonnet-5", "claude-sonnet-5-5", "deepseek-chat", "deepseek-reasoner", "deepseek-v4-flash", "deepseek-v4-pro", "fugu", "fugu-ultra", "fugu-ultra-20260615", "gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", "gemini-3.5-flash", "gemini-3.5-flash-lite", "gemini-3.6-flash", "gemini-3.7-flash", "gemini-3.8-flash", "gemini-4-argon", "gemma-4-26b-a4b-it", "gemma-4-31b", "gemma-4-31b-it", "glm-4.5", "glm-4.7", "glm-5", "glm-5-turbo", "glm-5.2", "glm-5.3", "glm-5.3-flash", "glm-5v-turbo", "gpt-5-codex", "gpt-5.2", "gpt-5.3-codex-spark", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-pro", "gpt-5.5", "gpt-5.6", "gpt-5.6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-6-astra", "gpt-6-luna", "gpt-6-sol", "gpt-6.1-sol", "gpt-oss-120b", "grok-4.3", "grok-4.6", "grok-4.7", "grok-build", "k3", "kilo-auto/small", "kimi-for-coding", "kimi-for-coding-highspeed", "kimi-k2.6", "kimi-k2.7-code", "kimi-k3", "kimi-latest", "lmstudio", "mimo-v2-flash", "mimo-v2.5", "mimo-v2.5-pro", "mimo-v2.5-pro-ultraspeed", "mimo-v2.6-flash", "mimo-v2.6-pro", "mimo-v2.6-pro-ultraspeed", "minimax-m3", "mistral-medium-latest", "mlx-community/Qwen3.6-27B-OptiQ-4bit", "muse-spark-1.2", "muse-spark-1.2-contributor", "omlx", "openai/gpt-oss-120b"];
export const TOOLS: ToolName[] = ["shell", "read_file", "edit_file", "write_file", "webfetch", "skill", "codedb", "read_tool_result", "render_html", "todo_write", "todo_read", "eval", "note_constraint", "ask_user", "attempt_completion", "load_tool_schemas", "mcp_search_tools", "mcp_select_tool", "clock_sleep", "subagent", "workflow", "agent_output", "agent_message", "subagent_resume", "learn_candidate", "peer_message", "workspace", "yield_turn", "imagegen", "rlm"];
diff --git a/spec/conformance.py b/spec/conformance.py
index b53261d4..09e831c7 100644
--- a/spec/conformance.py
+++ b/spec/conformance.py
@@ -312,8 +312,8 @@ def check_provider() -> int:
ids = [r["id"] for r in PROVIDER_SPECS]
if len(ids) != len(set(ids)):
raise Counterexample("provider-unique", None, f"ids={ids}")
- if len(PROVIDER_SPECS) != 24:
- raise Counterexample("provider-count", None, f"n={len(PROVIDER_SPECS)} want=24")
+ if len(PROVIDER_SPECS) != 25:
+ raise Counterexample("provider-count", None, f"n={len(PROVIDER_SPECS)} want=25")
responses = [r for r in PROVIDER_SPECS if r["kind"] == "responses"]
if [r["id"] for r in responses] != ["openai", "chatgpt-new", "codex"]:
raise Counterexample("responses-vendors", None, f"{responses}")
diff --git a/spec/kernels/providers.json b/spec/kernels/providers.json
index 5a5358b3..931fc05a 100644
--- a/spec/kernels/providers.json
+++ b/spec/kernels/providers.json
@@ -200,6 +200,15 @@
"sub_login": false,
"takes_effort": false
},
+ {
+ "id": "omlx",
+ "kind": "openai",
+ "auth": "bearer",
+ "login": "api_key",
+ "catalog": "openai",
+ "sub_login": false,
+ "takes_effort": true
+ },
{
"id": "chatgpt-new",
"kind": "responses",
diff --git a/spec/ref/provider.py b/spec/ref/provider.py
index 1ba8fb73..30c1f106 100644
--- a/spec/ref/provider.py
+++ b/spec/ref/provider.py
@@ -53,6 +53,7 @@ SPECS: list[dict] = [
row("fireworks", "openai", "bearer", catalog="openai"),
row("mlx", "openai", "bearer"),
row("lmstudio", "openai", "bearer"),
+ row("omlx", "openai", "bearer", catalog="openai", takes_effort=True),
row("chatgpt-new", "responses", "bearer", login="chatgpt_browser", catalog="openai", sub_login=True),
row("codex", "responses", "bearer", login="codex_device", catalog="codex", sub_login=True),
]
diff --git a/src/agent_request.zig b/src/agent_request.zig
index 69e9f204..cf7415f3 100644
--- a/src/agent_request.zig
+++ b/src/agent_request.zig
@@ -385,10 +385,12 @@ pub fn request(self: *Agent, tools_in: ?[]const u8) !std.json.ObjectMap {
}
continue;
}
- try self.say("[request failed: {t} — giving up this turn]\n", .{err});
+ // A refused loopback connection means the local server (oMLX, LM Studio, mlx-lm) is not running: name it.
+ const down: []const u8 = if (err == error.ConnectionRefused and @import("keys_cli.zig").isLocalUrl(self.provider.url)) self.provider.url else "";
+ try self.say("[request failed: {t}{s}{s} — giving up this turn]\n", .{ err, if (down.len > 0) " at " else "", down });
// Preserve whether this was provider throttling or a transport failure in the failed turn's JSON/trajectory (#86).
const failure_kind = if (err == error.RateLimited) "rate limited (429)" else if (err == error.ServerError) "server error (5xx)" else "network error";
- self.last_api_error = std.fmt.allocPrint(self.arena, "{s}: {s} (gave up after {d} attempts)", .{ failure_kind, @errorName(err), max_attempts }) catch null;
+ self.last_api_error = std.fmt.allocPrint(self.arena, "{s}: {s}{s}{s} (gave up after {d} attempts)", .{ failure_kind, @errorName(err), if (down.len > 0) " at " else "", down, max_attempts }) catch null;
self.last_request_write_failed = std.mem.eql(u8, @errorName(err), "WriteFailed");
if (telemetry.g_telem) |t| t.errorEvent("net", @errorName(err));
if (self.tracer) |tr| tr.api(self.label, self.sub, self.provider.model, 0, body.len, 0, 0, 0, true);
diff --git a/src/billing.zig b/src/billing.zig
index 1c3bedd1..bedcca2f 100644
--- a/src/billing.zig
+++ b/src/billing.zig
@@ -205,6 +205,7 @@ test "costFor: the picker's badge follows the credential, then the provider" {
// localhost is nobody's bill.
try std.testing.expectEqual(CostClass.local, costFor("lmstudio", .environment));
try std.testing.expectEqual(CostClass.local, costFor("mlx", .none));
+ try std.testing.expectEqual(CostClass.local, costFor("omlx", .stored));
// No credential: the row still says what signing in would buy.
try std.testing.expectEqual(CostClass.plan, costFor("codex", .none));
diff --git a/src/pricing.zig b/src/pricing.zig
index 7c5eac34..a212faa4 100644
--- a/src/pricing.zig
+++ b/src/pricing.zig
@@ -238,14 +238,14 @@ fn pureKimiGeneration(name: []const u8) ?u32 {
return std.fmt.parseInt(u32, name[1..], 10) catch null;
}
-/// Codex's first visible account model is its dynamic default. Kimi prefers
-/// the newest explicit pure generation (`k3`, then future `k4`, etc.) over
-/// compatibility aliases and K2 point-release ids.
+/// Codex's first visible account model is its dynamic default, and so is the
+/// first model oMLX lists. Kimi prefers the newest explicit pure generation
+/// (`k3`, then future `k4`, etc.) over compatibility aliases and K2 point-release ids.
pub fn providerDefaultModel(provider_id: []const u8, fallback: []const u8) []const u8 {
for (models()) |model| if (std.mem.eql(u8, model.provider, provider_id) and
std.mem.eql(u8, model.name, "mimo-v2.6-pro")) return model.name;
- if (std.mem.eql(u8, provider_id, "codex")) {
- for (models()) |model| if (std.mem.eql(u8, model.provider, "codex")) return model.name;
+ if (std.mem.eql(u8, provider_id, "codex") or std.mem.eql(u8, provider_id, "omlx")) {
+ for (models()) |model| if (std.mem.eql(u8, model.provider, provider_id)) return model.name;
return fallback;
}
if (!std.mem.eql(u8, provider_id, "kimi")) return fallback;
diff --git a/src/pricing_models.zig b/src/pricing_models.zig
index 4a703e51..52582c94 100644
--- a/src/pricing_models.zig
+++ b/src/pricing_models.zig
@@ -11,6 +11,9 @@ pub const rows = [_]ModelInfo{
// LM Studio serves whatever model is loaded; "lmstudio" is a routing alias —
// swap for your loaded model id if LM Studio requires an exact match (GET :1234/v1/models).
.{ .provider = "lmstudio", .name = "lmstudio", .context = 200_000 },
+ // oMLX needs an exact id. Its live /v1/models list replaces this offline alias
+ // and its first row becomes the default (pricing.providerDefaultModel).
+ .{ .provider = "omlx", .name = "omlx", .context = 200_000 },
// Anthropic: no-key/offline FALLBACK ONLY; with a key, router_catalog swaps
// in the live /v1/models list. Rows exist so boot, --schema and routing
// (the default, ladder and vision picks) resolve before that fetch.
diff --git a/src/provider.zig b/src/provider.zig
index 79a3c8f9..35600c16 100644
--- a/src/provider.zig
+++ b/src/provider.zig
@@ -117,6 +117,9 @@ pub const provider_specs = [_]ProviderSpec{
// lm-studio: the LM Studio app's local OpenAI-compatible server (default :1234).
// Load a model in LM Studio, then `LMSTUDIO_API_KEY=local graff --model lmstudio`.
.{ .id = "lmstudio", .display_name = "lmstudio", .kind = .openai, .auth = .bearer, .url = "http://127.0.0.1:1234/v1/chat/completions", .env_key = "LMSTUDIO_API_KEY", .default_model = "lmstudio" },
+ // omlx: the oMLX menu-bar app's local server (default :8000). It requires a bearer
+ // key, lists the downloaded models live, and takes reasoning_effort.
+ .{ .id = "omlx", .display_name = "oMLX", .kind = .openai, .auth = .bearer, .url = "http://127.0.0.1:8000/v1/chat/completions", .env_key = "OMLX_API_KEY", .default_model = "omlx", .catalog = .openai, .models_url = "http://127.0.0.1:8000/v1/models", .takes_effort = true },
// chatgpt-new: the person's ChatGPT plan through OpenAI's sign-in for
// open-source apps (ADR 0221). The key is the OAuth access token saved by
// `graff login chatgpt`, sent to the public Responses endpoint. Before
@@ -512,15 +515,6 @@ test "Keys.providerFor: known model, claude/gateway fallbacks, missing key" {
try std.testing.expectError(error.MissingKey, none.providerFor("claude-opus-4-8"));
}
-test "Keys.providerById: exact id wins, unknown id falls back to model routing" {
- const all = Keys{ .values = @splat("k") };
- const p = try all.providerById("anthropic", "claude-opus-4-8");
- try std.testing.expectEqualStrings("anthropic", p.id);
- try std.testing.expectEqualStrings("claude-opus-4-8", p.model);
- const fb = try all.providerById("no-such-provider", "gpt-5.5");
- try std.testing.expectEqualStrings("gpt-5.5", fb.model);
-}
-
test "Keys.defaultProvider: first keyed provider on its default model" {
const all = Keys{ .values = @splat("k") };
const p = try all.defaultProvider();
diff --git a/src/provider_tests.zig b/src/provider_tests.zig
index 299e1817..1ae98f41 100644
--- a/src/provider_tests.zig
+++ b/src/provider_tests.zig
@@ -8,6 +8,15 @@ const Keys = provider_mod.Keys;
const Provider = provider_mod.Provider;
const provider_specs = provider_mod.provider_specs;
+test "Keys.providerById: exact id wins, unknown id falls back to model routing" {
+ const all = Keys{ .values = @splat("k") };
+ const p = try all.providerById("anthropic", "claude-opus-4-8");
+ try std.testing.expectEqualStrings("anthropic", p.id);
+ try std.testing.expectEqualStrings("claude-opus-4-8", p.model);
+ const fb = try all.providerById("no-such-provider", "gpt-5.5");
+ try std.testing.expectEqualStrings("gpt-5.5", fb.model);
+}
+
test "Keys.build: g_codex_url_override rewires only the codex endpoint" {
const all = Keys{ .values = @splat("k") };
provider_mod.g_codex_url_override = "http://127.0.0.1:8765/responses";
diff --git a/src/router_catalog.zig b/src/router_catalog.zig
index f0a47c73..bf4e5cb2 100644
--- a/src/router_catalog.zig
+++ b/src/router_catalog.zig
@@ -191,10 +191,10 @@ pub fn parseModels(arena: Allocator, provider_id: []const u8, data: []const u8)
if (duplicate) continue;
var context = positiveInt(item.object, "context_length");
if (context == 0) context = positiveInt(item.object, "context");
- // Fireworks declares the window in camelCase (AIP gateway shape).
- if (context == 0) context = positiveInt(item.object, "contextLength");
+ if (context == 0) context = positiveInt(item.object, "contextLength"); // Fireworks (AIP gateway shape)
// Vercel AI Gateway (and some OpenAI-compat hosts) use context_window.
if (context == 0) context = positiveInt(item.object, "context_window");
+ if (context == 0) context = positiveInt(item.object, "max_model_len"); // oMLX and vLLM-style local servers
// Anthropic's input window is max_input_tokens; its sibling max_tokens is the
// OUTPUT cap (128k on Opus 5), never a window, or 1M would compact at ~102k.
if (context == 0) context = positiveInt(item.object, "max_input_tokens");
diff --git a/src/router_catalog_tests.zig b/src/router_catalog_tests.zig
index c192531c..03579ebd 100644
--- a/src/router_catalog_tests.zig
+++ b/src/router_catalog_tests.zig
@@ -145,6 +145,36 @@ test "parseModels is reusable for a newly configured router" {
try std.testing.expectEqual(pricing.default_context, snapshot.models[2].context);
}
+test "oMLX is a local live-catalog seat whose first listed model is the default" {
+ const spec = provider.specFor("omlx") orelse return error.TestUnexpectedResult;
+ try std.testing.expectEqual(provider.Provider.Kind.openai, spec.kind);
+ try std.testing.expectEqual(provider.Provider.Auth.bearer, spec.auth);
+ try std.testing.expectEqualStrings("http://127.0.0.1:8000/v1/chat/completions", spec.url);
+ try std.testing.expectEqualStrings("OMLX_API_KEY", spec.env_key);
+ try std.testing.expectEqual(provider.ProviderSpec.CatalogKind.openai, spec.catalog);
+ try std.testing.expectEqualStrings("http://127.0.0.1:8000/v1/models", modelsUrl(spec));
+ try std.testing.expect(spec.takes_effort);
+ // Offline (no catalog yet) the default is the baked alias.
+ try std.testing.expectEqualStrings("omlx", pricing.providerDefaultModel("omlx", spec.default_model));
+
+ var state = std.heap.ArenaAllocator.init(std.testing.allocator);
+ defer state.deinit();
+ const snapshot = parseModels(state.allocator(), "omlx",
+ \\{"object":"list","data":[
+ \\ {"id":"Local-27B-4bit","object":"model","owned_by":"omlx","max_model_len":262144},
+ \\ {"id":"Local-8B-4bit","object":"model","owned_by":"omlx"}
+ \\]}
+ ) orelse return error.TestUnexpectedResult;
+ try std.testing.expectEqual(@as(u64, 262_144), snapshot.models[0].context); // max_model_len read
+ try std.testing.expectEqual(pricing.default_context, snapshot.models[1].context);
+
+ const saved = pricing.active_model_table;
+ defer pricing.active_model_table = saved;
+ try std.testing.expect(activate(state.allocator(), spec, snapshot.models));
+ try std.testing.expectEqualStrings("Local-27B-4bit", pricing.providerDefaultModel("omlx", spec.default_model));
+ try std.testing.expect(!pricing.providerModelInTable("omlx", "omlx")); // the live list replaced the alias
+}
+
test "fireworks catalog is live and parses the AIP gateway shape" {
const spec = provider.specFor("fireworks") orelse return error.TestUnexpectedResult;
try std.testing.expectEqual(provider.ProviderSpec.CatalogKind.openai, spec.catalog);
diff --git a/src/schema.zig b/src/schema.zig
index db8e9ee7..ea77d2bc 100644
--- a/src/schema.zig
+++ b/src/schema.zig
@@ -557,6 +557,7 @@ test "providerTakesEffort: native xAI grok takes effort; gateway grok does not"
try std.testing.expect(providerTakesEffort(.openai, "zai", "glm-5.3"));
try std.testing.expect(providerTakesEffort(.openai, "vercel", "alibaba/qwen3.8-27b"));
try std.testing.expect(providerTakesEffort(.openai, "openrouter", "anthropic/claude-sonnet-4.6"));
+ try std.testing.expect(providerTakesEffort(.openai, "omlx", "any-downloaded-model"));
try std.testing.expect(providerTakesEffort(.openai, "kimi", "k3"));
try std.testing.expect(!providerTakesEffort(.openai, "openai", "gpt-5.5")); // direct openai chat
try std.testing.expect(providerTakesEffort(.responses, "xai", "grok-4.6"));