Skip to content

Agent Zero AI framework

v0.7 testing

frdele209a0c213fc
diff --git a/agent.py b/agent.py
index 5e4da261a..137f16198 100644
--- a/agent.py
+++ b/agent.py
@@ -1,7 +1,7 @@
import asyncio
from dataclasses import dataclass, field
import time, importlib, inspect, os, json
-from typing import Any, Optional, Dict
+from typing import Any, Optional, Dict, TypedDict
import uuid
from python.helpers import extract_tools, rate_limiter, files, errors
from python.helpers.print_style import PrintStyle
@@ -122,6 +122,15 @@ class AgentConfig:
additional: Dict[str, Any] = field(default_factory=dict)
+class LoopData:
+ def __init__(self):
+ self.iteration = -1
+ self.system = []
+ self.message = ""
+ self.history_from = 0
+ self.history = []
+
+
# intervention exception class - skips rest of message loop iteration
class InterventionException(Exception):
pass
@@ -162,40 +171,35 @@ class Agent:
async def monologue(self, msg: str):
try:
-
# loop data dictionary to pass to extensions
- loop_data: dict[str, Any] = {
- "message": msg,
- "iteration": -1,
- "history_from": len(self.history),
- }
+ loop_data = LoopData()
+ loop_data.message = msg
+ loop_data.history_from = len(self.history)
# call monologue_start extensions
await self.call_extensions("monologue_start", loop_data=loop_data)
printer = PrintStyle(italic=True, font_color="#b3ffd9", padding=False)
- user_message = self.read_prompt("fw.user_message.md", message=msg)
+ user_message = self.read_prompt("fw.user_message.md", message=loop_data.message)
await self.append_message(user_message, human=True)
- await self.call_extensions(
- "monologue_start", loop_data=loop_data
- ) # call monologue_end extensions
-
+
# let the agent run message loop until he stops it with a response tool
while True:
self.context.streaming_agent = self # mark self as current streamer
agent_response = ""
- loop_data["iteration"] += 1
+ loop_data.iteration += 1
try:
# set system prompt and message history
- loop_data["system"] = [
- self.read_prompt("agent.system.md", agent_name=self.agent_name),
- self.read_prompt("agent.system.tools.md"),
+ loop_data.system = [
+ self.read_prompt(
+ "agent.system.main.md", agent_name=self.agent_name
+ )
]
- loop_data["history"] = {"messages": self.history}
+ loop_data.history = self.history
# and allow extensions to edit them
await self.call_extensions(
@@ -205,7 +209,7 @@ class Agent:
# build chain from system prompt, message history and model
prompt = ChatPromptTemplate.from_messages(
[
- SystemMessage(content="\n\n".join(loop_data["system"])),
+ SystemMessage(content="\n\n".join(loop_data.system)),
MessagesPlaceholder(variable_name="messages"),
]
)
@@ -227,7 +231,7 @@ class Agent:
type="agent", heading=f"{self.agent_name}: Generating"
)
- async for chunk in chain.astream(loop_data["history"]):
+ async for chunk in chain.astream({"messages": loop_data.history}):
await self.handle_intervention(
agent_response
) # wait for intervention and handle it, if paused
@@ -324,6 +328,35 @@ class Agent:
)
return content
+ def read_prompts(self, pattern: str, **kwargs):
+ import glob
+
+ prompts = []
+
+ # Scan both configured subdir and default folder
+ subdir_files = glob.glob(
+ files.get_abs_path("prompts", self.config.prompts_subdir, pattern)
+ )
+ default_files = glob.glob(files.get_abs_path("prompts", "default", pattern))
+
+ # Create a dictionary to store files, prioritizing the config subdir
+ files_to_read = {file.split("/")[-1]: file for file in default_files}
+
+ # Override with files from subdir if they exist
+ for file in subdir_files:
+ files_to_read[file.split("/")[-1]] = file
+
+ # Sort files alphabetically by their file names
+ sorted_files = sorted(files_to_read.items())
+
+ # Read the files in alphabetical order
+ for filename, filepath in sorted_files:
+ content = files.read_file(files.get_abs_path(filepath), **kwargs)
+ if content:
+ prompts.append(content)
+
+ return prompts
+
def get_data(self, field: str):
return self.data.get(field, None)
diff --git a/default/index.faiss b/default/index.faiss
new file mode 100644
index 000000000..6e6feb72f
Binary files /dev/null and b/default/index.faiss differ
diff --git a/default/index.pkl b/default/index.pkl
new file mode 100644
index 000000000..2014ef12e
Binary files /dev/null and b/default/index.pkl differ
diff --git a/initialize.py b/initialize.py
index cafe43130..177c102a0 100644
--- a/initialize.py
+++ b/initialize.py
@@ -12,7 +12,7 @@ def initialize():
# chat_llm = models.get_anthropic_chat(model_name="claude-3-5-sonnet-20240620", temperature=0)
# chat_llm = models.get_google_chat(model_name="gemini-1.5-flash", temperature=0)
# chat_llm = models.get_mistral_chat(model_name="mistral-small-latest", temperature=0)
- # chat_llm = models.get_groq_chat(model_name="llama-3.1-70b-versatile", temperature=0)
+ # chat_llm = models.get_groq_chat(model_name="llama-3.2-90b-text-preview", temperature=0)
# chat_llm = models.get_sambanova_chat(model_name="Meta-Llama-3.1-70B-Instruct-8k", temperature=0)
# utility model used for helper functions (cheaper, faster)
@@ -35,7 +35,7 @@ def initialize():
auto_memory_count = 0,
# auto_memory_skip = 2,
# rate_limit_seconds = 60,
- rate_limit_requests = 15,
+ rate_limit_requests = 30,
# rate_limit_input_tokens = 0,
# rate_limit_output_tokens = 0,
# msgs_keep_max = 25,
diff --git a/prompts/default/agent.system.md b/prompts/default/agent.system.main.md
similarity index 100%
rename from prompts/default/agent.system.md
rename to prompts/default/agent.system.main.md
diff --git a/prompts/default/agent.system.solutions.md b/prompts/default/agent.system.solutions.md
index 8b8eb8997..0d8743d15 100644
--- a/prompts/default/agent.system.solutions.md
+++ b/prompts/default/agent.system.solutions.md
@@ -1,4 +1,4 @@
-# Solutions in the past
+# Solutions from the past
- following are your memories about successful solutions of related problems:
{{solutions}}
\ No newline at end of file
diff --git a/prompts/default/agent.system.tool.call_sub.md b/prompts/default/agent.system.tool.call_sub.md
new file mode 100644
index 000000000..ac055e2ed
--- /dev/null
+++ b/prompts/default/agent.system.tool.call_sub.md
@@ -0,0 +1,20 @@
+### call_subordinate:
+Use subordinate agents to solve subtasks.
+Use "message" argument to send message. Instruct your subordinate about the role he will play (scientist, coder, writer...) and his task in detail.
+Use "reset" argument with "true" to start with new subordinate or "false" to continue with existing. For brand new tasks use "true", for followup conversation use "false".
+Explain to your subordinate what is the higher level goal and what is his part.
+Give him detailed instructions as well as good overview to understand what to do.
+**Example usage**:
+~~~json
+{
+ "thoughts": [
+ "The result seems to be ok but...",
+ "I will ask my subordinate to fix...",
+ ],
+ "tool_name": "call_subordinate",
+ "tool_args": {
+ "message": "Well done, now edit...",
+ "reset": "false"
+ }
+}
+~~~
\ No newline at end of file
diff --git a/prompts/default/agent.system.tool.code_exe.md b/prompts/default/agent.system.tool.code_exe.md
new file mode 100644
index 000000000..a71a772b6
--- /dev/null
+++ b/prompts/default/agent.system.tool.code_exe.md
@@ -0,0 +1,84 @@
+### code_execution_tool:
+Execute provided terminal commands, python code or nodejs code.
+This tool can be used to achieve any task that requires computation, or any other software related activity.
+Place your code escaped and properly indented in the "code" argument.
+Select the corresponding runtime with "runtime" argument. Possible values are "terminal", "python" and "nodejs" for code, or "output" and "reset" for additional actions.
+Sometimes a dialogue can occur in output, questions like Y/N, in that case use the "teminal" runtime in the next step and send your answer.
+If the code is running long, you can use runtime "output" to wait for next output part or use runtime "reset" to kill the process.
+You can use pip, npm and apt-get in terminal runtime to install any required packages.
+IMPORTANT: Never use implicit print or implicit output, it does not work! If you need output of your code, you MUST use print() or console.log() to output selected variables.
+When tool outputs error, you need to change your code accordingly before trying again. knowledge_tool can help analyze errors.
+IMPORTANT!: Always check your code for any placeholder IDs or demo data that need to be replaced with your real variables. Do not simply reuse code snippets from tutorials.
+Do not use in combination with other tools except for thoughts. Wait for response before using other tools.
+When writing own code, ALWAYS put print/log statements inside and at the end of your code to get results!
+**Example usages:**
+1. Execute python code
+~~~json
+{
+ "thoughts": [
+ "I need to do...",
+ "I can use library...",
+ "Then I can...",
+ ],
+ "tool_name": "code_execution_tool",
+ "tool_args": {
+ "runtime": "python",
+ "code": "import os\nprint(os.getcwd())",
+ }
+}
+~~~
+
+2. Execute terminal command
+~~~json
+{
+ "thoughts": [
+ "I need to do...",
+ "I need to install...",
+ ],
+ "tool_name": "code_execution_tool",
+ "tool_args": {
+ "runtime": "terminal",
+ "code": "apt-get install zip",
+ }
+}
+~~~
+
+2. 1. Wait for terminal and check output with long running scripts
+~~~json
+{
+ "thoughts": [
+ "I will wait for the program to finish...",
+ ],
+ "tool_name": "code_execution_tool",
+ "tool_args": {
+ "runtime": "output",
+ }
+}
+~~~
+
+2. 2. Answer terminal dialog
+~~~json
+{
+ "thoughts": [
+ "Program needs confirmation...",
+ ],
+ "tool_name": "code_execution_tool",
+ "tool_args": {
+ "runtime": "terminal",
+ "code": "Y",
+ }
+}
+~~~
+
+2. 3. Reset terminal
+~~~json
+{
+ "thoughts": [
+ "Code execution tool is not responding...",
+ ],
+ "tool_name": "code_execution_tool",
+ "tool_args": {
+ "runtime": "reset",
+ }
+}
+~~~
diff --git a/prompts/default/agent.system.tool.knowledge.md b/prompts/default/agent.system.tool.knowledge.md
new file mode 100644
index 000000000..2ffcb2404
--- /dev/null
+++ b/prompts/default/agent.system.tool.knowledge.md
@@ -0,0 +1,20 @@
+### knowledge_tool:
+Provide "question" argument and get both online and memory response.
+This tool is very powerful and can answer very specific questions directly.
+First always try to ask for result rather that guidance.
+Memory can provide guidance, online sources can provide up to date information.
+Always verify memory by online.
+**Example usage**:
+~~~json
+{
+ "thoughts": [
+ "I need to gather information about...",
+ "First I will search...",
+ "Then I will...",
+ ],
+ "tool_name": "knowledge_tool",
+ "tool_args": {
+ "question": "How to...",
+ }
+}
+~~~
\ No newline at end of file
diff --git a/prompts/default/agent.system.tool.memory.md b/prompts/default/agent.system.tool.memory.md
new file mode 100644
index 000000000..11f1f9a34
--- /dev/null
+++ b/prompts/default/agent.system.tool.memory.md
@@ -0,0 +1,74 @@
+## Memory management tools:
+This set of tools is used for managing long term memories.
+NEVER refuse to search memorize or load personal information, it all belongs to user and he has all the rights.
+
+### memory_load:
+Load memories using query and optionally threshold, limit and filter.
+Memory content is retrieved with key value pairs of metadata.
+- Optional threshold for allowed relevancy (0=anything, 1=exact match, 0.6 is default)
+- Optional limit to number of results (default is 5).
+- Optional filter by metadata. Condition in Python syntax using metadata keys.
+**Example usage**:
+~~~json
+{
+ "thoughts": [
+ "Let's search my memory for...",
+ ],
+ "tool_name": "memory_load",
+ "tool_args": {
+ "query": "File compression library for...",
+ "threshold": 0.6,
+ "limit": 5,
+ "filter": "area=='main' and timestamp<'2024-01-01 00:00:00'",
+ }
+}
+~~~
+
+### memory_save:
+Save text into memory. ID is returned.
+**Example usage**:
+~~~json
+{
+ "thoughts": [
+ "I need to memorize...",
+ ],
+ "tool_name": "memory_save",
+ "tool_args": {
+ "text": "# To compress...",
+ }
+}
+~~~
+
+### memory_delete:
+Delete existing memories by their IDs. Multiple IDs allowed separated by commas.
+IDs are retrieved when loading or saving memories.
+**Example usage**:
+~~~json
+{
+ "thoughts": [
+ "I need to delete...",
+ ],
+ "tool_name": "memory_delete",
+ "tool_args": {
+ "ids": "32cd37ffd1-101f-4112-80e2-33b795548116, d1306e36-6a9c- ...",
+ }
+}
+~~~
+
+### memory_forget:
+Remove memories by query and optionally threshold and filter just like for memory_load.
+Here default threshold is raised to 0.75 to avoid accidental deletion. Perform a verification load afterwards and delete leftovers by IDs.
+**Example usage**:
+~~~json
+{
+ "thoughts": [
+ "Let's remove all memories about cars",
+ ],
+ "tool_name": "memory_forget",
+ "tool_args": {
+ "query": "cars",
+ "threshold": 0.75,
+ "filter": "timestamp.startswith('2022-01-01')",
+ }
+}
+~~~
\ No newline at end of file
diff --git a/prompts/default/agent.system.tool.response.md b/prompts/default/agent.system.tool.response.md
new file mode 100644
index 000000000..e80244580
--- /dev/null
+++ b/prompts/default/agent.system.tool.response.md
@@ -0,0 +1,19 @@
+### response:
+Final answer for user.
+Ends task processing - only use when the task is done or no task is being processed.
+Place your result in "text" argument.
+Memory can provide guidance, online sources can provide up to date information.
+Always verify memory by online.
+**Example usage**:
+~~~json
+{
+ "thoughts": [
+ "The user has greeted me...",
+ "I will...",
+ ],
+ "tool_name": "response",
+ "tool_args": {
+ "text": "Hi...",
+ }
+}
+~~~
\ No newline at end of file
diff --git a/prompts/default/agent.system.tool.web.md b/prompts/default/agent.system.tool.web.md
new file mode 100644
index 000000000..a4fc49c51
--- /dev/null
+++ b/prompts/default/agent.system.tool.web.md
@@ -0,0 +1,19 @@
+### webpage_content_tool:
+Retrieves the text content of a webpage, such as a news article or Wikipedia page.
+Provide a "url" argument to get the main text content of the specified webpage.
+This tool is useful for gathering information from online sources.
+Always provide a full, valid URL including the protocol (http:// or https://).
+
+**Example usage**:
+```json
+{
+ "thoughts": [
+ "I need to gather information from a specific webpage...",
+ "I will use the webpage_content_tool to fetch the content...",
+ ],
+ "tool_name": "webpage_content_tool",
+ "tool_args": {
+ "url": "https://en.wikipedia.org/wiki/Artificial_intelligence",
+ }
+}
+```
\ No newline at end of file
diff --git a/prompts/default/agent.system.tools.md b/prompts/default/agent.system.tools.md
index 4b2d7b76c..34ca96951 100644
--- a/prompts/default/agent.system.tools.md
+++ b/prompts/default/agent.system.tools.md
@@ -1,230 +1,3 @@
## Tools available:
-### response:
-Final answer for user.
-Ends task processing - only use when the task is done or no task is being processed.
-Place your result in "text" argument.
-Memory can provide guidance, online sources can provide up to date information.
-Always verify memory by online.
-**Example usage**:
-~~~json
-{
- "thoughts": [
- "The user has greeted me...",
- "I will...",
- ],
- "tool_name": "response",
- "tool_args": {
- "text": "Hi...",
- }
-}
-~~~
-
-### call_subordinate:
-Use subordinate agents to solve subtasks.
-Use "message" argument to send message. Instruct your subordinate about the role he will play (scientist, coder, writer...) and his task in detail.
-Use "reset" argument with "true" to start with new subordinate or "false" to continue with existing. For brand new tasks use "true", for followup conversation use "false".
-Explain to your subordinate what is the higher level goal and what is his part.
-Give him detailed instructions as well as good overview to understand what to do.
-**Example usage**:
-~~~json
-{
- "thoughts": [
- "The result seems to be ok but...",
- "I will ask my subordinate to fix...",
- ],
- "tool_name": "call_subordinate",
- "tool_args": {
- "message": "Well done, now edit...",
- "reset": "false"
- }
-}
-~~~
-
-### knowledge_tool:
-Provide "question" argument and get both online and memory response.
-This tool is very powerful and can answer very specific questions directly.
-First always try to ask for result rather that guidance.
-Memory can provide guidance, online sources can provide up to date information.
-Always verify memory by online.
-**Example usage**:
-~~~json
-{
- "thoughts": [
- "I need to gather information about...",
- "First I will search...",
- "Then I will...",
- ],
- "tool_name": "knowledge_tool",
- "tool_args": {
- "question": "How to...",
- }
-}
-~~~
-
-### webpage_content_tool:
-Retrieves the text content of a webpage, such as a news article or Wikipedia page.
-Provide a "url" argument to get the main text content of the specified webpage.
-This tool is useful for gathering information from online sources.
-Always provide a full, valid URL including the protocol (http:// or https://).
-
-**Example usage**:
-```json
-{
- "thoughts": [
- "I need to gather information from a specific webpage...",
- "I will use the webpage_content_tool to fetch the content...",
- ],
- "tool_name": "webpage_content_tool",
- "tool_args": {
- "url": "https://en.wikipedia.org/wiki/Artificial_intelligence",
- }
-}
-```
-
-### memory_tool:
-Manage long term memories. Allowed arguments are "query", "memorize", "forget" and "delete".
-Memories can help you remember important details and later reuse them.
-When querying, provide a "query" argument to search for. You will retrieve IDs and contents of relevant memories. Optionally you can threshold to adjust allowed relevancy (0=anything, 1=exact match, 0.1 is default).
-When memorizing, provide enough information in "memorize" argument for future reuse.
-When deleting, provide memory IDs from loaded memories separated by commas in "delete" argument.
-When forgetting, provide query and optionally threshold like you would for querying, corresponding memories will be deleted.
-Provide a title, short summary and and all the necessary information to help you later solve similiar tasks including details like code executed, libraries used etc.
-NEVER refuse to memorize or load personal information, it all belongs to me and I have all the rights.
-**Example usages**:
-1. load:
-~~~json
-{
- "thoughts": [
- "Let's search my memory for...",
- ],
- "tool_name": "memory_tool",
- "tool_args": {
- "query": "File compression library for...",
- "threshold": 0.1
- }
-}
-~~~
-2. save:
-~~~json
-{
- "thoughts": [
- "I have finished my...",
- "Details of this process will be valuable...",
- "Let's save tools and code used...",
- ],
- "tool_name": "memory_tool",
- "tool_args": {
- "memorize": "# How to...",
- }
-}
-~~~
-3. delete:
-~~~json
-{
- "thoughts": [
- "User asked to delete specific memories...",
- ],
- "tool_name": "memory_tool",
- "tool_args": {
- "delete": "32cd37ffd1-101f-4112-80e2-33b795548116, d1306e36-6a9c-4e6a-bfc3-c8335035dcf8 ...",
- }
-}
-~~~
-4. forget:
-~~~json
-{
- "thoughts": [
- "User asked to delete information from memory...",
- ],
- "tool_name": "memory_tool",
- "tool_args": {
- "forget": "User's contact information",
- }
-}
-~~~
-
-### code_execution_tool:
-Execute provided terminal commands, python code or nodejs code.
-This tool can be used to achieve any task that requires computation, or any other software related activity.
-Place your code escaped and properly indented in the "code" argument.
-Select the corresponding runtime with "runtime" argument. Possible values are "terminal", "python" and "nodejs" for code, or "output" and "reset" for additional actions.
-Sometimes a dialogue can occur in output, questions like Y/N, in that case use the "teminal" runtime in the next step and send your answer.
-If the code is running long, you can use runtime "output" to wait for next output part or use runtime "reset" to kill the process.
-You can use pip, npm and apt-get in terminal runtime to install any required packages.
-IMPORTANT: Never use implicit print or implicit output, it does not work! If you need output of your code, you MUST use print() or console.log() to output selected variables.
-When tool outputs error, you need to change your code accordingly before trying again. knowledge_tool can help analyze errors.
-IMPORTANT!: Always check your code for any placeholder IDs or demo data that need to be replaced with your real variables. Do not simply reuse code snippets from tutorials.
-Do not use in combination with other tools except for thoughts. Wait for response before using other tools.
-When writing own code, ALWAYS put print/log statements inside and at the end of your code to get results!
-**Example usages:**
-1. Execute python code
-~~~json
-{
- "thoughts": [
- "I need to do...",
- "I can use library...",
- "Then I can...",
- ],
- "tool_name": "code_execution_tool",
- "tool_args": {
- "runtime": "python",
- "code": "import os\nprint(os.getcwd())",
- }
-}
-~~~
-
-2. Execute terminal command
-~~~json
-{
- "thoughts": [
- "I need to do...",
- "I need to install...",
- ],
- "tool_name": "code_execution_tool",
- "tool_args": {
- "runtime": "terminal",
- "code": "apt-get install zip",
- }
-}
-~~~
-
-2. 1. Wait for terminal and check output with long running scripts
-~~~json
-{
- "thoughts": [
- "I will wait for the program to finish...",
- ],
- "tool_name": "code_execution_tool",
- "tool_args": {
- "runtime": "output",
- }
-}
-~~~
-
-2. 2. Answer terminal dialog
-~~~json
-{
- "thoughts": [
- "Program needs confirmation...",
- ],
- "tool_name": "code_execution_tool",
- "tool_args": {
- "runtime": "terminal",
- "code": "Y",
- }
-}
-~~~
-
-2. 3. Reset terminal
-~~~json
-{
- "thoughts": [
- "Code execution tool is not responding...",
- ],
- "tool_name": "code_execution_tool",
- "tool_args": {
- "runtime": "reset",
- }
-}
-~~~
+{{tools}}
\ No newline at end of file
diff --git a/prompts/default/fw.memory_saved.md b/prompts/default/fw.memory_saved.md
index 73e21b4d6..ea9ddb0dc 100644
--- a/prompts/default/fw.memory_saved.md
+++ b/prompts/default/fw.memory_saved.md
@@ -1,5 +1 @@
-~~~json
-{
- "memory": "Memory has been saved with id {{memory_id}}."
-}
-~~~
\ No newline at end of file
+Memory saved with id {{memory_id}}
\ No newline at end of file
diff --git a/prompts/default/fw.msg_truncated.md b/prompts/default/fw.msg_truncated.md
index b7237bd4a..cc8597fee 100644
--- a/prompts/default/fw.msg_truncated.md
+++ b/prompts/default/fw.msg_truncated.md
@@ -1 +1 @@
-<< REMOVED TO SAVE SPACE >>
\ No newline at end of file
+<< {{length}} CHARACTERS REMOVED TO SAVE SPACE >>
\ No newline at end of file
diff --git a/python/extensions/message_loop_prompts/_10_tool_instructions.py b/python/extensions/message_loop_prompts/_10_tool_instructions.py
new file mode 100644
index 000000000..89de8d898
--- /dev/null
+++ b/python/extensions/message_loop_prompts/_10_tool_instructions.py
@@ -0,0 +1,23 @@
+from python.helpers.extension import Extension
+from agent import Agent, LoopData
+
+
+class RecallMemories(Extension):
+
+ INTERVAL = 3
+ HISTORY = 5
+ RESULTS = 3
+ THRESHOLD = 0.1
+
+ async def execute(self, loop_data: LoopData = LoopData(), **kwargs):
+ # collect and concatenate tool instructions
+ sys = concat_tool_prompts(self.agent)
+ # append to system message
+ loop_data.system.append(sys)
+
+
+def concat_tool_prompts(agent: Agent):
+ tools = agent.read_prompts("agent.system.tool.*.md")
+ tools = "\n\n".join(tools)
+ sys = agent.read_prompt("agent.system.tools.md", tools=tools)
+ return sys
diff --git a/python/extensions/message_loop_prompts/_50_recall_memories.py b/python/extensions/message_loop_prompts/_50_recall_memories.py
new file mode 100644
index 000000000..a3a1cb951
--- /dev/null
+++ b/python/extensions/message_loop_prompts/_50_recall_memories.py
@@ -0,0 +1,92 @@
+from python.helpers.extension import Extension
+from python.helpers.memory import Memory
+from agent import LoopData
+
+
+class RecallMemories(Extension):
+
+ INTERVAL = 3
+ HISTORY = 5
+ RESULTS = 3
+ THRESHOLD = 0.1
+
+ async def execute(self, loop_data: LoopData = LoopData(), **kwargs):
+
+ if (
+ loop_data.iteration % RecallMemories.INTERVAL == 0
+ ): # every 3 iterations (or the first one) recall memories
+ await self.search_memories(loop_data=loop_data, **kwargs)
+
+ async def search_memories(self, loop_data: LoopData, **kwargs):
+ # try:
+ # show temp info message
+ self.agent.context.log.log(
+ type="info", content="Searching memories...", temp=True
+ )
+
+ # show full util message, this will hide temp message immediately if turned on
+ log_item = self.agent.context.log.log(
+ type="util",
+ heading="Searching memories...",
+ )
+
+ # get system message and chat history for util llm
+ msgs_text = self.agent.concat_messages(
+ self.agent.history[-RecallMemories.HISTORY :]
+ ) # only last X messages
+ system = self.agent.read_prompt(
+ "memory.memories_query.sys.md", history=msgs_text
+ )
+
+ # log query streamed by LLM
+ def log_callback(content):
+ log_item.stream(query=content)
+
+ # call util llm to summarize conversation
+ query = await self.agent.call_utility_llm(
+ system=system, msg=loop_data.message, callback=log_callback
+ )
+
+ # get solutions database
+ db = await Memory.get(self.agent)
+
+ memories = await db.search_similarity_threshold(
+ query=query,
+ limit=RecallMemories.RESULTS,
+ threshold=RecallMemories.THRESHOLD,
+ filter=f"area != '{Memory.Area.SOLUTIONS.value}'", # exclude solutions
+ )
+
+ # log the short result
+ if not isinstance(memories, list) or len(memories) == 0:
+ log_item.update(
+ heading="No useful memories found.",
+ )
+ return
+ else:
+ log_item.update(
+ heading=f"\n\n{len(memories)} memories found.",
+ )
+
+ # concatenate memory.page_content in memories:
+ memories_text = ""
+ for memory in memories:
+ memories_text += memory.page_content + "\n\n"
+ memories_text = memories_text.strip()
+
+ # log the full results
+ log_item.update(memories=memories_text)
+
+ # place to prompt
+ memories_prompt = self.agent.read_prompt(
+ "agent.system.memories.md", memories=memories_text
+ )
+
+ # append to system message
+ loop_data.system.append(memories_prompt)
+
+ # except Exception as e:
+ # err = errors.format_error(e)
+ # self.agent.context.log.log(
+ # type="error", heading="Recall memories extension error:", content=err
+ # )
diff --git a/python/extensions/message_loop_prompts/recall_solutions.py b/python/extensions/message_loop_prompts/_51_recall_solutions.py
similarity index 78%
rename from python/extensions/message_loop_prompts/recall_solutions.py
rename to python/extensions/message_loop_prompts/_51_recall_solutions.py
index 60f3997ed..d05c2500f 100644
--- a/python/extensions/message_loop_prompts/recall_solutions.py
+++ b/python/extensions/message_loop_prompts/_51_recall_solutions.py
@@ -1,10 +1,6 @@
-from agent import Agent
from python.helpers.extension import Extension
-from python.helpers.files import read_file
-from python.helpers.vector_db import Area
-import json
-from python.helpers import errors, files
-from python.tools.memory_tool import get_db
+from python.helpers.memory import Memory
+from agent import LoopData
class RecallSolutions(Extension):
@@ -14,15 +10,14 @@ class RecallSolutions(Extension):
RESULTS = 3
THRESHOLD = 0.1
- async def execute(self, loop_data={}, **kwargs):
+ async def execute(self, loop_data: LoopData = LoopData(), **kwargs):
- iter = loop_data.get("iteration", 0)
if (
- iter % RecallSolutions.INTERVAL == 0
+ loop_data.iteration % RecallSolutions.INTERVAL == 0
): # every 3 iterations (or the first one) recall solution memories
await self.search_solutions(loop_data=loop_data, **kwargs)
- async def search_solutions(self, loop_data={}, **kwargs):
+ async def search_solutions(self, loop_data: LoopData, **kwargs):
# try:
# show temp info message
self.agent.context.log.log(
@@ -49,17 +44,17 @@ class RecallSolutions(Extension):
# call util llm to summarize conversation
query = await self.agent.call_utility_llm(
- system=system, msg=loop_data["message"], callback=log_callback
+ system=system, msg=loop_data.message, callback=log_callback
)
# get solutions database
- vdb = get_db(self.agent)
+ db = await Memory.get(self.agent)
- solutions = vdb.search_similarity_threshold(
+ solutions = await db.search_similarity_threshold(
query=query,
- results=RecallSolutions.RESULTS,
+ limit=RecallSolutions.RESULTS,
threshold=RecallSolutions.THRESHOLD,
- filter=f"area == '{Area.SOLUTIONS.value}'"
+ filter=f"area == '{Memory.Area.SOLUTIONS.value}'"
)
# log the short result
@@ -88,7 +83,7 @@ class RecallSolutions(Extension):
)
# append to system message
- loop_data["system"] += solutions_prompt
+ loop_data.system.append(solutions_prompt)
# except Exception as e:
# err = errors.format_error(e)
diff --git a/python/extensions/message_loop_prompts/recall_memories.py b/python/extensions/message_loop_prompts/recall_memories.py
deleted file mode 100644
index a14092f2e..000000000
--- a/python/extensions/message_loop_prompts/recall_memories.py
+++ /dev/null
@@ -1,97 +0,0 @@
-from agent import Agent
-from python.helpers.extension import Extension
-from python.helpers.files import read_file
-from python.helpers.vector_db import Area
-import json
-from python.helpers import errors, files
-from python.tools.memory_tool import get_db
-
-
-class RecallMemories(Extension):
-
- INTERVAL = 3
- HISTORY = 5
- RESULTS = 3
- THRESHOLD = 0.1
-
- async def execute(self, loop_data={}, **kwargs):
-
- iter = loop_data.get("iteration", 0)
- if (
- iter % RecallMemories.INTERVAL == 0
- ): # every 3 iterations (or the first one) recall memories
- await self.search_memories(loop_data=loop_data, **kwargs)
-
- async def search_memories(self, loop_data={}, **kwargs):
- # try:
- # show temp info message
- self.agent.context.log.log(
- type="info", content="Searching memories...", temp=True
- )
-
- # show full util message, this will hide temp message immediately if turned on
- log_item = self.agent.context.log.log(
- type="util",
- heading="Searching memories...",
- )
-
- # get system message and chat history for util llm
- msgs_text = self.agent.concat_messages(
- self.agent.history[-RecallMemories.HISTORY :]
- ) # only last X messages
- system = self.agent.read_prompt(
- "memory.memories_query.sys.md", history=msgs_text
- )
-
- # log query streamed by LLM
- def log_callback(content):
- log_item.stream(query=content)
-
- # call util llm to summarize conversation
- query = await self.agent.call_utility_llm(
- system=system, msg=loop_data["message"], callback=log_callback
- )
-
- # get solutions database
- vdb = get_db(self.agent)
-
- memories = vdb.search_similarity_threshold(
- query=query,
- results=RecallMemories.RESULTS,
- threshold=RecallMemories.THRESHOLD,
- filter=f"area != '{Area.SOLUTIONS.value}'" # exclude solutions
- )
-
- # log the short result
- if not isinstance(memories, list) or len(memories) == 0:
- log_item.update(
- heading="No useful memories found.",
- )
- return
- else:
- log_item.update(
- heading=f"\n\n{len(memories)} memories found.",
- )
-
- # concatenate memory.page_content in memories:
- memories_text = ""
- for memory in memories:
- memories_text += memory.page_content + "\n\n"
- memories_text = memories_text.strip()
-
- # log the full results
- log_item.update(memories=memories_text)
-
- # place to prompt
- memories_prompt = self.agent.read_prompt(
- "agent.system.memories.md", memories=memories_text
- )
-
- # append to system message
- loop_data["system"] += memories_prompt
-
- # except Exception as e:
- # err = errors.format_error(e)
- # self.agent.context.log.log(
- # type="error", heading="Recall memories extension error:", content=err
- # )
diff --git a/python/extensions/monologue_end/50_memorize_memories.py b/python/extensions/monologue_end/50_memorize_memories.py
deleted file mode 100644
index 6a4c7dfa2..000000000
--- a/python/extensions/monologue_end/50_memorize_memories.py
+++ /dev/null
@@ -1,74 +0,0 @@
-from agent import Agent
-from python.helpers.extension import Extension
-import python.helpers.files as files
-from python.helpers.vector_db import Area
-import json
-from python.helpers.dirty_json import DirtyJson
-from python.helpers import errors
-from python.tools.memory_tool import get_db
-
-class MemorizeMemories(Extension):
-
- async def execute(self, loop_data={}, **kwargs):
- # try:
-
- # show temp info message
- self.agent.context.log.log(
- type="info", content="Memorizing new information...", temp=True
- )
-
- # show full util message, this will hide temp message immediately if turned on
- log_item = self.agent.context.log.log(
- type="util",
- heading="Memorizing new information...",
- )
-
- # get system message and chat history for util llm
- system = self.agent.read_prompt("memory.memories_sum.sys.md")
- msgs_text = self.agent.concat_messages(self.agent.history)
-
- # log query streamed by LLM
- def log_callback(content):
- log_item.stream(content=content)
-
- # call util llm to find info in history
- memories_json = await self.agent.call_utility_llm(
- system=system,
- msg=msgs_text,
- callback=log_callback,
- )
-
- memories = DirtyJson.parse_string(memories_json)
-
- if not isinstance(memories, list) or len(memories) == 0:
- log_item.update(heading="No useful information to memorize.")
- return
- else:
- log_item.update(
- heading=f"{len(memories)} entries to memorize."
- )
-
- # save chat history
- vdb = get_db(self.agent)
-
- memories_txt = ""
- for memory in memories:
- # solution to plain text:
- txt = f"{memory}"
- memories_txt += txt + "\n\n"
- vdb.insert_text(
- text=txt, metadata={"area": Area.MAIN.value}
- )
-
- memories_txt = memories_txt.strip()
- log_item.update(memories=memories_txt)
- log_item.update(
- result=f"{len(memories)} entries memorized.",
- heading=f"{len(memories)} entries memorized.",
- )
-
- # except Exception as e:
- # err = errors.format_error(e)
- # self.agent.context.log.log(
- # type="error", heading="Memorize memories extension error:", content=err
- # )
diff --git a/python/extensions/monologue_end/51_memorize_solutions.py b/python/extensions/monologue_end/51_memorize_solutions.py
deleted file mode 100644
index 459ab0dde..000000000
--- a/python/extensions/monologue_end/51_memorize_solutions.py
+++ /dev/null
@@ -1,74 +0,0 @@
-from agent import Agent
-from python.helpers.extension import Extension
-import python.helpers.files as files
-from python.helpers.vector_db import Area
-import json
-from python.helpers.dirty_json import DirtyJson
-from python.helpers import errors
-from python.tools.memory_tool import get_db
-
-class MemorizeSolutions(Extension):
-
- async def execute(self, loop_data={}, **kwargs):
- # try:
-
- # show temp info message
- self.agent.context.log.log(
- type="info", content="Memorizing succesful solutions...", temp=True
- )
-
- # show full util message, this will hide temp message immediately if turned on
- log_item = self.agent.context.log.log(
- type="util",
- heading="Memorizing succesful solutions...",
- )
-
- # get system message and chat history for util llm
- system = self.agent.read_prompt("memory.solutions_sum.sys.md")
- msgs_text = self.agent.concat_messages(self.agent.history)
-
- # log query streamed by LLM
- def log_callback(content):
- log_item.stream(content=content)
-
- # call util llm to find solutions in history
- solutions_json = await self.agent.call_utility_llm(
- system=system,
- msg=msgs_text,
- callback=log_callback,
- )
-
- solutions = DirtyJson.parse_string(solutions_json)
-
- if not isinstance(solutions, list) or len(solutions) == 0:
- log_item.update(heading="No successful solutions to memorize.")
- return
- else:
- log_item.update(
- heading=f"{len(solutions)} successful solutions to memorize."
- )
-
- # save chat history
- vdb = get_db(self.agent)
-
- solutions_txt = ""
- for solution in solutions:
- # solution to plain text:
- txt = f"# Problem\n {solution['problem']}\n# Solution\n {solution['solution']}"
- solutions_txt += txt + "\n\n"
- vdb.insert_text(
- text=txt, metadata={"area": Area.SOLUTIONS.value}
- )
-
- solutions_txt = solutions_txt.strip()
- log_item.update(solutions=solutions_txt)
- log_item.update(
- result=f"{len(solutions)} solutions memorized.",
- heading=f"{len(solutions)} solutions memorized.",
- )
-
- # except Exception as e:
- # err = errors.format_error(e)
- # self.agent.context.log.log(
- # type="error", heading="Memorize solutions extension error:", content=err
- # )
diff --git a/python/extensions/monologue_end/_50_memorize_memories.py b/python/extensions/monologue_end/_50_memorize_memories.py
new file mode 100644
index 000000000..7781c4e1e
--- /dev/null
+++ b/python/extensions/monologue_end/_50_memorize_memories.py
@@ -0,0 +1,94 @@
+import asyncio
+from python.helpers.extension import Extension
+from python.helpers.memory import Memory
+from python.helpers.dirty_json import DirtyJson
+from agent import LoopData
+from python.helpers.log import LogItem
+from python.helpers.defer import run_in_background
+
+
+
+class MemorizeMemories(Extension):
+
+ REPLACE_THRESHOLD = 0.9
+
+ async def execute(self, loop_data: LoopData = LoopData(), **kwargs):
+ # try:
+
+ # show temp info message
+ self.agent.context.log.log(
+ type="info", content="Memorizing new information...", temp=True
+ )
+
+ # show full util message, this will hide temp message immediately if turned on
+ log_item = self.agent.context.log.log(
+ type="util",
+ heading="Memorizing new information...",
+ )
+
+ #memorize in background
+ asyncio.create_task(self.memorize(loop_data, log_item))
+
+ async def memorize(self, loop_data: LoopData, log_item: LogItem, **kwargs):
+
+ # get system message and chat history for util llm
+ system = self.agent.read_prompt("memory.memories_sum.sys.md")
+ msgs_text = self.agent.concat_messages(self.agent.history)
+
+ # log query streamed by LLM
+ def log_callback(content):
+ log_item.stream(content=content)
+
+ # call util llm to find info in history
+ memories_json = await self.agent.call_utility_llm(
+ system=system,
+ msg=msgs_text,
+ callback=log_callback,
+ )
+
+ memories = DirtyJson.parse_string(memories_json)
+
+ if not isinstance(memories, list) or len(memories) == 0:
+ log_item.update(heading="No useful information to memorize.")
+ return
+ else:
+ log_item.update(heading=f"{len(memories)} entries to memorize.")
+
+ # save chat history
+ db = await Memory.get(self.agent)
+
+ memories_txt = ""
+ rem = []
+ for memory in memories:
+ # solution to plain text:
+ txt = f"{memory}"
+ memories_txt += "\n\n" + txt
+ log_item.update(memories=memories_txt.strip())
+
+ # remove previous solutions too similiar to this one
+ if self.REPLACE_THRESHOLD > 0:
+ rem += await db.delete_documents_by_query(
+ query=txt,
+ threshold=self.REPLACE_THRESHOLD,
+ filter=f"area=='{Memory.Area.MAIN.value}'",
+ )
+ rem_txt = "\n\n".join(Memory.format_docs_plain(rem))
+ log_item.update(replaced=rem_txt)
+
+
+
+ # insert new solution
+ db.insert_text(text=txt, metadata={"area": Memory.Area.MAIN.value})
+
+ log_item.update(
+ result=f"{len(memories)} entries memorized.",
+ heading=f"{len(memories)} entries memorized.",
+ )
+ if rem:
+ log_item.stream(result=f"\nReplaced {len(rem)} previous memories.")
+
+ # except Exception as e:
+ # err = errors.format_error(e)
+ # self.agent.context.log.log(
+ # type="error", heading="Memorize memories extension error:", content=err
+ # )
diff --git a/python/extensions/monologue_end/_51_memorize_solutions.py b/python/extensions/monologue_end/_51_memorize_solutions.py
new file mode 100644
index 000000000..2f86b9bea
--- /dev/null
+++ b/python/extensions/monologue_end/_51_memorize_solutions.py
@@ -0,0 +1,92 @@
+import asyncio
+from python.helpers.extension import Extension
+from python.helpers.memory import Memory
+from python.helpers.dirty_json import DirtyJson
+from agent import LoopData
+from python.helpers.log import LogItem
+
+
+class MemorizeSolutions(Extension):
+
+ REPLACE_THRESHOLD = 0.9
+
+ async def execute(self, loop_data: LoopData = LoopData(), **kwargs):
+ # try:
+
+ # show temp info message
+ self.agent.context.log.log(
+ type="info", content="Memorizing succesful solutions...", temp=True
+ )
+
+ # show full util message, this will hide temp message immediately if turned on
+ log_item = self.agent.context.log.log(
+ type="util",
+ heading="Memorizing succesful solutions...",
+ )
+
+ #memorize in background
+ asyncio.create_task(self.memorize(loop_data, log_item))
+
+ async def memorize(self, loop_data: LoopData, log_item: LogItem, **kwargs):
+ # get system message and chat history for util llm
+ system = self.agent.read_prompt("memory.solutions_sum.sys.md")
+ msgs_text = self.agent.concat_messages(self.agent.history)
+
+ # log query streamed by LLM
+ def log_callback(content):
+ log_item.stream(content=content)
+
+ # call util llm to find solutions in history
+ solutions_json = await self.agent.call_utility_llm(
+ system=system,
+ msg=msgs_text,
+ callback=log_callback,
+ )
+
+ solutions = DirtyJson.parse_string(solutions_json)
+
+ if not isinstance(solutions, list) or len(solutions) == 0:
+ log_item.update(heading="No successful solutions to memorize.")
+ return
+ else:
+ log_item.update(
+ heading=f"{len(solutions)} successful solutions to memorize."
+ )
+
+ # save chat history
+ db = await Memory.get(self.agent)
+
+ solutions_txt = ""
+ rem = []
+ for solution in solutions:
+ # solution to plain text:
+ txt = f"# Problem\n {solution['problem']}\n# Solution\n {solution['solution']}"
+ solutions_txt += txt + "\n\n"
+
+ # remove previous solutions too similiar to this one
+ if self.REPLACE_THRESHOLD > 0:
+ rem += await db.delete_documents_by_query(
+ query=txt,
+ threshold=self.REPLACE_THRESHOLD,
+ filter=f"area=='{Memory.Area.SOLUTIONS.value}'",
+ )
+ rem_txt = "\n\n".join(Memory.format_docs_plain(rem))
+ log_item.update(replaced=rem_txt)
+
+ # insert new solution
+ db.insert_text(text=txt, metadata={"area": Memory.Area.SOLUTIONS.value})
+
+ solutions_txt = solutions_txt.strip()
+ log_item.update(solutions=solutions_txt)
+ log_item.update(
+ result=f"{len(solutions)} solutions memorized.",
+ heading=f"{len(solutions)} solutions memorized.",
+ )
+ if rem:
+ log_item.stream(result=f"\nReplaced {len(rem)} previous solutions.")
+
+ # except Exception as e:
+ # err = errors.format_error(e)
+ # self.agent.context.log.log(
+ # type="error", heading="Memorize solutions extension error:", content=err
+ # )
diff --git a/python/extensions/monologue_end/90_waiting_for_input_msg.py b/python/extensions/monologue_end/_90_waiting_for_input_msg.py
similarity index 73%
rename from python/extensions/monologue_end/90_waiting_for_input_msg.py
rename to python/extensions/monologue_end/_90_waiting_for_input_msg.py
index 2d3b558e8..448bc9017 100644
--- a/python/extensions/monologue_end/90_waiting_for_input_msg.py
+++ b/python/extensions/monologue_end/_90_waiting_for_input_msg.py
@@ -1,9 +1,9 @@
-from agent import Agent
from python.helpers.extension import Extension
+from agent import LoopData
class WaitingForInputMsg(Extension):
- async def execute(self, loop_data={}, **kwargs):
+ async def execute(self, loop_data: LoopData = LoopData(), **kwargs):
# show temp info message
if self.agent.number == 0:
self.agent.context.log.log(
diff --git a/python/helpers/defer.py b/python/helpers/defer.py
index 9151a7479..8446e5df9 100644
--- a/python/helpers/defer.py
+++ b/python/helpers/defer.py
@@ -1,6 +1,6 @@
import asyncio
import threading
-from concurrent.futures import Future
+from concurrent.futures import Future, ThreadPoolExecutor
from typing import Any, Callable, Optional, Coroutine
class EventLoopThread:
@@ -71,4 +71,10 @@ class DeferredTask:
return self._future and not self._future.done() # type: ignore
def restart(self) -> None:
- self._start_task()
\ No newline at end of file
+ self._start_task()
+
+def run_in_background(func, *args, **kwargs):
+ async def wrapper(*args, **kwargs):
+ loop = asyncio.get_event_loop()
+ return await loop.run_in_executor(None, func, *args, **kwargs)
+ return wrapper
\ No newline at end of file
diff --git a/python/helpers/extract_tools.py b/python/helpers/extract_tools.py
index f60655072..e061ebfc2 100644
--- a/python/helpers/extract_tools.py
+++ b/python/helpers/extract_tools.py
@@ -51,22 +51,30 @@ def fix_json_string(json_string):
T = TypeVar('T') # Define a generic type variable
def load_classes_from_folder(folder: str, name_pattern: str, base_class: Type[T]) -> list[Type[T]]:
+ import os
+ import importlib
+ import inspect
+ from fnmatch import fnmatch
classes = []
- # Get all .py files in the folder that match the pattern
- for file_name in os.listdir(folder):
- if fnmatch(file_name, name_pattern) and file_name.endswith(".py"):
- module_name = file_name[:-3] # remove .py extension
- module_path = folder.replace(os.sep, ".") + "." + module_name
- module = importlib.import_module(module_path)
+ # Get all .py files in the folder that match the pattern, sorted alphabetically
+ py_files = sorted(
+ [file_name for file_name in os.listdir(folder) if fnmatch(file_name, name_pattern) and file_name.endswith(".py")]
+ )
- # Get all classes in thde module
- class_list = inspect.getmembers(module, inspect.isclass)
+ # Iterate through the sorted list of files
+ for file_name in py_files:
+ module_name = file_name[:-3] # remove .py extension
+ module_path = folder.replace(os.sep, ".") + "." + module_name
+ module = importlib.import_module(module_path)
- # Filter for classes that are subclasses of the given base_class
- for cls in class_list:
- if cls[1] is not base_class and issubclass(cls[1], base_class):
- classes.append(cls[1])
+ # Get all classes in the module
+ class_list = inspect.getmembers(module, inspect.isclass)
- return classes
\ No newline at end of file
+ # Filter for classes that are subclasses of the given base_class
+ for cls in class_list:
+ if cls[1] is not base_class and issubclass(cls[1], base_class):
+ classes.append(cls[1])
+
+ return classes
diff --git a/python/helpers/knowledge_import.py b/python/helpers/knowledge_import.py
index 54f2152a7..7c2e65d0d 100644
--- a/python/helpers/knowledge_import.py
+++ b/python/helpers/knowledge_import.py
@@ -36,9 +36,9 @@ def calculate_checksum(file_path: str) -> str:
def load_knowledge(
log_item: LogItem | None, knowledge_dir: str, index: Dict[str, KnowledgeImport]
) -> Dict[str, KnowledgeImport]:
- knowledge_dir = files.get_abs_path(knowledge_dir)
+ knowledge_dir = files.get_abs_path("knowledge",knowledge_dir)
- from python.helpers.vector_db import Area
+ from python.helpers.memory import Memory
# Mapping file extensions to corresponding loader classes
file_types_loaders = {
@@ -53,7 +53,7 @@ def load_knowledge(
cnt_files = 0
cnt_docs = 0
- for area in Area:
+ for area in Memory.Area:
subdir = files.get_abs_path(knowledge_dir, area.value)
if not os.path.exists(subdir):
diff --git a/python/helpers/log.py b/python/helpers/log.py
index 1c1efe836..03eb63fc8 100644
--- a/python/helpers/log.py
+++ b/python/helpers/log.py
@@ -81,6 +81,7 @@ class Log:
self.updates: list[int] = []
self.logs: list[LogItem] = []
self.progress = ""
+ self.progress_no = 0
def log(
self,
@@ -101,8 +102,9 @@ class Log:
)
self.logs.append(item)
self.updates += [item.no]
- if heading:
+ if heading and item.no >= self.progress_no:
self.progress = heading
+ self.progress_no = item.no
return item
def update_item(
@@ -120,7 +122,9 @@ class Log:
item.type = type
if heading is not None:
item.heading = heading
- self.progress = heading
+ if no >= self.progress_no:
+ self.progress = heading
+ self.progress_no = no
if content is not None:
item.content = content
if kvps is not None:
@@ -156,3 +160,4 @@ class Log:
self.updates = []
self.logs = []
self.progress = ""
+ self.progress_no = 0
diff --git a/python/helpers/memory.py b/python/helpers/memory.py
new file mode 100644
index 000000000..6fa9b6026
--- /dev/null
+++ b/python/helpers/memory.py
@@ -0,0 +1,306 @@
+from datetime import datetime
+from typing import Any
+from langchain.storage import InMemoryByteStore, LocalFileStore
+from langchain.embeddings import CacheBackedEmbeddings
+
+# from langchain_chroma import Chroma
+from langchain_community.vectorstores import FAISS
+import faiss
+from langchain_community.docstore.in_memory import InMemoryDocstore
+from langchain_community.vectorstores.utils import (
+ DistanceStrategy,
+)
+import os, json
+
+import numpy as np
+from . import files
+from langchain_core.documents import Document
+import uuid
+from python.helpers import knowledge_import
+from python.helpers.log import Log, LogItem
+from enum import Enum
+from agent import Agent
+
+
+class Memory:
+
+ class Area(Enum):
+ MAIN = "main"
+ SOLUTIONS = "solutions"
+
+ index: dict[str, "FAISS"] = {}
+
+ @staticmethod
+ async def get(agent: Agent):
+ memory_subdir = agent.config.memory_subdir or "default"
+ if Memory.index.get(memory_subdir) is None:
+ log_item = agent.context.log.log(
+ type="util",
+ heading=f"Initializing VectorDB in '/{memory_subdir}'",
+ )
+ db = Memory.initialize(
+ log_item,
+ agent.config.embeddings_model,
+ memory_subdir,
+ False,
+ )
+ Memory.index[memory_subdir] = db
+ wrap = Memory(agent, db, memory_subdir=memory_subdir)
+ if agent.config.knowledge_subdirs:
+ await wrap.preload_knowledge(
+ log_item, agent.config.knowledge_subdirs, memory_subdir
+ )
+ return wrap
+ else:
+ return Memory(
+ agent=agent,
+ db=Memory.index[memory_subdir],
+ memory_subdir=memory_subdir,
+ )
+
+ @staticmethod
+ def initialize(
+ log_item: LogItem | None,
+ embeddings_model,
+ memory_subdir: str,
+ in_memory=False,
+ ):
+
+ print("Initializing VectorDB...")
+
+ if log_item:
+ log_item.stream(progress="\nInitializing VectorDB")
+
+ em_dir = files.get_abs_path(
+ "memory/embeddings"
+ ) # just caching, no need to parameterize
+ db_dir = Memory._abs_db_dir(memory_subdir)
+
+ # make sure embeddings and database directories exist
+ os.makedirs(db_dir, exist_ok=True)
+
+ if in_memory:
+ store = InMemoryByteStore()
+ else:
+ os.makedirs(em_dir, exist_ok=True)
+ store = LocalFileStore(em_dir)
+
+ # here we setup the embeddings model with the chosen cache storage
+ embedder = CacheBackedEmbeddings.from_bytes_store(
+ embeddings_model,
+ store,
+ namespace=getattr(
+ embeddings_model,
+ "model",
+ getattr(embeddings_model, "model_name", "default"),
+ ),
+ )
+
+ # self.db = Chroma(
+ # embedding_function=self.embedder,
+ # persist_directory=db_dir)
+
+ # if db folder exists and is not empty:
+ if os.path.exists(db_dir) and files.exists(db_dir, "index.faiss"):
+ db = FAISS.load_local(
+ folder_path=db_dir,
+ embeddings=embedder,
+ allow_dangerous_deserialization=True,
+ distance_strategy=DistanceStrategy.COSINE,
+ # normalize_L2=True,
+ relevance_score_fn=Memory._cosine_normalizer,
+ )
+ else:
+ index = faiss.IndexFlatIP(len(embedder.embed_query("example")))
+
+ db = FAISS(
+ embedding_function=embedder,
+ index=index,
+ docstore=InMemoryDocstore(),
+ index_to_docstore_id={},
+ distance_strategy=DistanceStrategy.COSINE,
+ # normalize_L2=True,
+ relevance_score_fn=Memory._cosine_normalizer,
+ )
+ return db
+
+ def __init__(
+ self,
+ agent: Agent,
+ db: FAISS,
+ memory_subdir: str,
+ ):
+ self.agent = agent
+ self.db = db
+ self.memory_subdir = memory_subdir
+
+ async def preload_knowledge(
+ self, log_item: LogItem | None, kn_dirs: list[str], memory_subdir: str
+ ):
+ # db abs path
+ db_dir = Memory._abs_db_dir(memory_subdir)
+
+ # Load the index file if it exists
+ index_path = files.get_abs_path(db_dir, "knowledge_import.json")
+
+ # make sure directory exists
+ if not os.path.exists(db_dir):
+ os.makedirs(db_dir)
+
+ index: dict[str, knowledge_import.KnowledgeImport] = {}
+ if os.path.exists(index_path):
+ with open(index_path, "r") as f:
+ index = json.load(f)
+
+ for kn_dir in kn_dirs:
+ index = knowledge_import.load_knowledge(log_item, kn_dir, index)
+
+ for file in index:
+ if index[file]["state"] in ["changed", "removed"] and index[file].get(
+ "ids", []
+ ): # for knowledge files that have been changed or removed and have IDs
+ await self.delete_documents_by_ids(
+ index[file]["ids"]
+ ) # remove original version
+ if index[file]["state"] == "changed":
+ index[file]["ids"] = self.insert_documents(
+ index[file]["documents"]
+ ) # insert new version
+
+ # remove index where state="removed"
+ index = {k: v for k, v in index.items() if v["state"] != "removed"}
+
+ # strip state and documents from index and save it
+ for file in index:
+ if "documents" in index[file]:
+ del index[file]["documents"] # type: ignore
+ if "state" in index[file]:
+ del index[file]["state"] # type: ignore
+ with open(index_path, "w") as f:
+ json.dump(index, f)
+
+ async def search_similarity_threshold(
+ self, query: str, limit: int, threshold: float, filter: str = ""
+ ):
+ comparator = Memory._get_comparator(filter) if filter else None
+ return await self.db.asearch(
+ query,
+ search_type="similarity_score_threshold",
+ k=limit,
+ score_threshold=threshold,
+ filter=comparator,
+ )
+
+ async def delete_documents_by_query(
+ self, query: str, threshold: float, filter: str = ""
+ ):
+ k = 100
+ tot = 0
+ removed = []
+
+ while True:
+ # Perform similarity search with score
+ docs = await self.search_similarity_threshold(
+ query, limit=k, threshold=threshold, filter=filter
+ )
+ removed += docs
+
+ # Extract document IDs and filter based on score
+ # document_ids = [result[0].metadata["id"] for result in docs if result[1] < score_limit]
+ document_ids = [result.metadata["id"] for result in docs]
+
+ # Delete documents with IDs over the threshold score
+ if document_ids:
+ # fnd = self.db.get(where={"id": {"$in": document_ids}})
+ # if fnd["ids"]: self.db.delete(ids=fnd["ids"])
+ # tot += len(fnd["ids"])
+ self.db.delete(ids=document_ids)
+ tot += len(document_ids)
+
+ # If fewer than K document IDs, break the loop
+ if len(document_ids) < k:
+ break
+
+ if tot:
+ self._save_db() # persist
+ return removed
+
+ async def delete_documents_by_ids(self, ids: list[str]):
+ # pre = self.db.get(ids=ids)["ids"]
+ self.db.delete(ids=ids)
+ # post = self.db.get(ids=ids)["ids"]
+ # TODO? compare pre and post
+ if ids:
+ self._save_db() # persist
+ return len(ids)
+
+ def insert_text(self, text, metadata: dict = {}):
+ id = str(uuid.uuid4())
+ self.db.add_documents(
+ documents=[
+ Document(
+ text,
+ metadata={"id": id, "timestamp": self.get_timestamp(), **metadata},
+ )
+ ],
+ ids=[id],
+ )
+ self._save_db() # persist
+ return id
+
+ def insert_documents(self, docs: list[Document]):
+ ids = [str(uuid.uuid4()) for _ in range(len(docs))]
+ timestamp = self.get_timestamp()
+ if ids:
+ for doc, id in zip(docs, ids):
+ doc.metadata["id"] = id # add ids to documents metadata
+ doc.metadata["timestamp"] = timestamp # add timestamp
+ self.db.add_documents(documents=docs, ids=ids)
+ self._save_db() # persist
+ return ids
+
+ def _save_db(self):
+ self.db.save_local(folder_path=self._abs_db_dir(self.memory_subdir))
+
+ @staticmethod
+ def _get_comparator(condition: str):
+ def comparator(data: dict[str, Any]):
+ try:
+ return eval(condition, {}, data)
+ except Exception as e:
+ # print(f"Error evaluating condition: {e}")
+ return False
+
+ return comparator
+
+ @staticmethod
+ def _score_normalizer(val: float) -> float:
+ res = 1 - 1 / (1 + np.exp(val))
+ return res
+
+ @staticmethod
+ def _cosine_normalizer(val: float) -> float:
+ res = (1 + val) / 2
+ res = max(
+ 0, min(1, res)
+ ) # float precision can cause values like 1.0000000596046448
+ return res
+
+ @staticmethod
+ def _abs_db_dir(memory_subdir: str) -> str:
+ return files.get_abs_path("memory", memory_subdir)
+
+ @staticmethod
+ def format_docs_plain(docs: list[Document]) -> list[str]:
+ result = []
+ for doc in docs:
+ text = ""
+ for k, v in doc.metadata.items():
+ text += f"{k}: {v}\n"
+ text += f"Content: {doc.page_content}"
+ result.append(text)
+ return result
+
+ @staticmethod
+ def get_timestamp():
+ return datetime.now().strftime("%Y-%m-%d %H:%M:%S")
diff --git a/python/helpers/messages.py b/python/helpers/messages.py
index fceccb132..6033e2532 100644
--- a/python/helpers/messages.py
+++ b/python/helpers/messages.py
@@ -5,8 +5,8 @@ def truncate_text(agent, output, threshold=1000):
return output
# Adjust the file path as needed
- placeholder = agent.read_prompt("fw.msg_truncated.md", removed_chars=(len(output) - threshold))
- # placeholder = files.read_file("./prompts/default/fw.msg_truncated.md", removed_chars=(len(output) - threshold))
+ placeholder = agent.read_prompt("fw.msg_truncated.md", length=(len(output) - threshold))
+ # placeholder = files.read_file("./prompts/default/fw.msg_truncated.md", length=(len(output) - threshold))
start_len = (threshold - len(placeholder)) // 2
end_len = threshold - len(placeholder) - start_len
diff --git a/python/helpers/vector_db.py b/python/helpers/vector_db.py
deleted file mode 100644
index 2456393a3..000000000
--- a/python/helpers/vector_db.py
+++ /dev/null
@@ -1,234 +0,0 @@
-from typing import Any
-from langchain.storage import InMemoryByteStore, LocalFileStore
-from langchain.embeddings import CacheBackedEmbeddings
-
-# from langchain_chroma import Chroma
-from langchain_community.vectorstores import FAISS
-import faiss
-from langchain_community.docstore.in_memory import InMemoryDocstore
-
-import os, json
-from . import files
-from langchain_core.documents import Document
-import uuid
-from python.helpers import knowledge_import
-from python.helpers.log import Log, LogItem
-import pandas as pd
-from enum import Enum
-
-
-class Area(Enum):
- MAIN = "main"
- SOLUTIONS = "solutions"
-
-
-index: dict[str, "VectorDB"] = {}
-
-
-def get_or_create_db(
- logger: Log | None,
- embeddings_model,
- memory_dir: str,
- in_memory=False,
- knowledge_dirs: list[str] = [],
-):
- if index.get(memory_dir) is None:
- log_item = None
- if(logger): log_item = logger.log(type="util", heading=f"Initializing VectorDB in {memory_dir}")
- index[memory_dir] = VectorDB(
- log_item, embeddings_model, memory_dir, knowledge_dirs, in_memory
- )
- return index[memory_dir]
-
-
-class VectorDB:
-
- def __init__(
- self,
- log_item: LogItem | None,
- embeddings_model,
- memory_dir: str,
- knowledge_dirs: list[str] = [],
- in_memory=False,
- ):
- self.log_item = log_item
-
- print("Initializing VectorDB...")
- if(self.log_item): self.log_item.stream(progress="\nInitializing VectorDB")
-
- self.embeddings_model = embeddings_model
-
- self.em_dir = files.get_abs_path(
- "./memory/embeddings"
- ) # just caching, no need to parameterize
- self.db_dir = files.get_abs_path("./memory", memory_dir, "database")
-
- # make sure embeddings and database directories exist
- os.makedirs(self.db_dir, exist_ok=True)
-
- if in_memory:
- self.store = InMemoryByteStore()
- else:
- os.makedirs(self.em_dir, exist_ok=True)
- self.store = LocalFileStore(self.em_dir)
-
- # here we setup the embeddings model with the chosen cache storage
- self.embedder = CacheBackedEmbeddings.from_bytes_store(
- embeddings_model,
- self.store,
- namespace=getattr(
- embeddings_model,
- "model",
- getattr(embeddings_model, "model_name", "default"),
- ),
- )
-
- # self.db = Chroma(
- # embedding_function=self.embedder,
- # persist_directory=db_dir)
-
- # if db folder exists and is not empty:
- if os.path.exists(self.db_dir) and files.exists(self.db_dir, "index.faiss"):
- self.db = FAISS.load_local(
- folder_path=self.db_dir,
- embeddings=self.embedder,
- allow_dangerous_deserialization=True,
- )
- else:
- index = faiss.IndexFlatL2(len(self.embedder.embed_query("example text")))
-
- self.db = FAISS(
- embedding_function=self.embedder,
- index=index,
- docstore=InMemoryDocstore(),
- index_to_docstore_id={},
- )
-
- # preload knowledge files
- if knowledge_dirs:
- self.preload_knowledge(knowledge_dirs, self.db_dir)
-
- def preload_knowledge(self, kn_dirs: list[str], db_dir: str):
-
- # Load the index file if it exists
- index_path = files.get_abs_path(db_dir, "knowledge_import.json")
-
- # make sure directory exists
- if not os.path.exists(db_dir):
- os.makedirs(db_dir)
-
- index: dict[str, knowledge_import.KnowledgeImport] = {}
- if os.path.exists(index_path):
- with open(index_path, "r") as f:
- index = json.load(f)
-
- for kn_dir in kn_dirs:
- index = knowledge_import.load_knowledge(self.log_item, kn_dir, index)
-
- for file in index:
- if index[file]["state"] in ["changed", "removed"] and index[file].get(
- "ids", []
- ): # for knowledge files that have been changed or removed and have IDs
- self.delete_documents_by_ids(
- index[file]["ids"]
- ) # remove original version
- if index[file]["state"] == "changed":
- index[file]["ids"] = self.insert_documents(
- index[file]["documents"]
- ) # insert new version
-
- # remove index where state="removed"
- index = {k: v for k, v in index.items() if v["state"] != "removed"}
-
- # strip state and documents from index and save it
- for file in index:
- if "documents" in index[file]:
- del index[file]["documents"] # type: ignore
- if "state" in index[file]:
- del index[file]["state"] # type: ignore
- with open(index_path, "w") as f:
- json.dump(index, f)
-
- def search_similarity(self, query, results=3):
- return self.db.similarity_search(query, results)
-
- def search_similarity_threshold(
- self, query: str, results=3, threshold=0.5, filter: str = ""
- ):
- comparator = VectorDB.get_comparator(filter) if filter else None
- return self.db.search(
- query,
- search_type="similarity_score_threshold",
- k=results,
- score_threshold=threshold,
- filter=comparator,
- )
-
- def search_max_rel(self, query, results=3):
- return self.db.max_marginal_relevance_search(query, results)
-
- def delete_documents_by_query(self, query: str, threshold=0.1):
- k = 100
- tot = 0
- while True:
- # Perform similarity search with score
- docs = self.search_similarity_threshold(
- query, results=k, threshold=threshold
- )
-
- # Extract document IDs and filter based on score
- # document_ids = [result[0].metadata["id"] for result in docs if result[1] < score_limit]
- document_ids = [result.metadata["id"] for result in docs]
-
- # Delete documents with IDs over the threshold score
- if document_ids:
- # fnd = self.db.get(where={"id": {"$in": document_ids}})
- # if fnd["ids"]: self.db.delete(ids=fnd["ids"])
- # tot += len(fnd["ids"])
- self.db.delete(ids=document_ids)
- tot += len(document_ids)
-
- # If fewer than K document IDs, break the loop
- if len(document_ids) < k:
- break
-
- if tot:
- self.db.save_local(folder_path=self.db_dir) # persist
- return tot
-
- def delete_documents_by_ids(self, ids: list[str]):
- # pre = self.db.get(ids=ids)["ids"]
- self.db.delete(ids=ids)
- # post = self.db.get(ids=ids)["ids"]
- # TODO? compare pre and post
- if ids:
- self.db.save_local(folder_path=self.db_dir) # persist
- return len(ids)
-
- def insert_text(self, text, metadata: dict = {}):
- id = str(uuid.uuid4())
- self.db.add_documents(
- documents=[Document(text, metadata={"id": id, **metadata})], ids=[id]
- )
- self.db.save_local(folder_path=self.db_dir) # persist
- return id
-
- def insert_documents(self, docs: list[Document]):
- ids = [str(uuid.uuid4()) for _ in range(len(docs))]
- if ids:
- for doc, id in zip(docs, ids):
- doc.metadata["id"] = id # add ids to documents metadata
- self.db.add_documents(documents=docs, ids=ids)
- self.db.save_local(folder_path=self.db_dir) # persist
- return ids
-
- @staticmethod
- def get_comparator(condition: str):
- def comparator(data: dict[str, Any]):
- try:
- return eval(condition, {}, data)
- except Exception as e:
- print(f"Error evaluating condition: {e}")
- return False
-
- return comparator
diff --git a/python/tools/call_subordinate.py b/python/tools/call_subordinate.py
index d754590a0..227e87a03 100644
--- a/python/tools/call_subordinate.py
+++ b/python/tools/call_subordinate.py
@@ -10,4 +10,5 @@ class Delegation(Tool):
subordinate.set_data("superior", self.agent)
self.agent.set_data("subordinate", subordinate)
# run subordinate agent message loop
- return Response( message= await self.agent.get_data("subordinate").message_loop(message), break_loop=False)
\ No newline at end of file
+ subordinate: Agent = self.agent.get_data("subordinate")
+ return Response( message= await subordinate.monologue(message), break_loop=False)
\ No newline at end of file
diff --git a/python/tools/knowledge_tool.py b/python/tools/knowledge_tool.py
index b0e9029da..d4b09744f 100644
--- a/python/tools/knowledge_tool.py
+++ b/python/tools/knowledge_tool.py
@@ -1,55 +1,56 @@
import os
-from python.helpers import perplexity_search
-from python.helpers import duckduckgo_search
-from . import memory_tool
-import concurrent.futures
+import asyncio
+from python.helpers import memory, perplexity_search, duckduckgo_search
from python.helpers.tool import Tool, Response
from python.helpers.print_style import PrintStyle
from python.helpers.errors import handle_error
class Knowledge(Tool):
async def execute(self, question="", **kwargs):
- with concurrent.futures.ThreadPoolExecutor() as executor:
- # Schedule the two functions to be run in parallel
-
- # perplexity search, if API key provided
- if os.getenv("API_KEY_PERPLEXITY"):
- perplexity = executor.submit(perplexity_search.perplexity_search, question)
- else:
- PrintStyle.hint("No API key provided for Perplexity. Skipping Perplexity search.")
- self.agent.context.log.log(type="hint", content="No API key provided for Perplexity. Skipping Perplexity search.")
- perplexity = None
-
-
- # duckduckgo search
- duckduckgo = executor.submit(duckduckgo_search.search, question)
-
- # manual memory search
- future_memory_man = executor.submit(memory_tool.search, self.agent, question)
-
- # Wait for both functions to complete
- try:
- perplexity_result = (perplexity.result() if perplexity else "") or ""
- except Exception as e:
- handle_error(e)
- perplexity_result = "Perplexity search failed: " + str(e)
-
- try:
- duckduckgo_result = duckduckgo.result()
- except Exception as e:
- handle_error(e)
- duckduckgo_result = "DuckDuckGo search failed: " + str(e)
-
- try:
- memory_result = future_memory_man.result()
- except Exception as e:
- handle_error(e)
- memory_result = "Memory search failed: " + str(e)
+ # Create tasks for all three search methods
+ tasks = [
+ self.perplexity_search(question),
+ self.duckduckgo_search(question),
+ self.mem_search(question)
+ ]
+
+ # Run all tasks concurrently
+ results = await asyncio.gather(*tasks, return_exceptions=True)
+
+ perplexity_result, duckduckgo_result, memory_result = results
+
+ # Handle exceptions and format results
+ perplexity_result = self.format_result(perplexity_result, "Perplexity")
+ duckduckgo_result = self.format_result(duckduckgo_result, "DuckDuckGo")
+ memory_result = self.format_result(memory_result, "Memory")
msg = self.agent.read_prompt("tool.knowledge.response.md",
- online_sources = ((perplexity_result + "\n\n") if perplexity else "") + str(duckduckgo_result),
- memory = memory_result )
+ online_sources = ((perplexity_result + "\n\n") if perplexity_result else "") + str(duckduckgo_result),
+ memory = memory_result)
- await self.agent.handle_intervention(msg) # wait for intervention and handle it, if paused
+ await self.agent.handle_intervention(msg) # wait for intervention and handle it, if paused
return Response(message=msg, break_loop=False)
+
+ async def perplexity_search(self, question):
+ if os.getenv("API_KEY_PERPLEXITY"):
+ return await asyncio.to_thread(perplexity_search.perplexity_search, question)
+ else:
+ PrintStyle.hint("No API key provided for Perplexity. Skipping Perplexity search.")
+ self.agent.context.log.log(type="hint", content="No API key provided for Perplexity. Skipping Perplexity search.")
+ return None
+
+ async def duckduckgo_search(self, question):
+ return await asyncio.to_thread(duckduckgo_search.search, question)
+
+ async def mem_search(self, question: str):
+ db = await memory.Memory.get(self.agent)
+ docs = await db.search_similarity_threshold(query=question, limit=5, threshold=0.5)
+ text = memory.Memory.format_docs_plain(docs)
+ return "\n\n".join(text)
+
+ def format_result(self, result, source):
+ if isinstance(result, Exception):
+ handle_error(result)
+ return f"{source} search failed: {str(result)}"
+ return result if result else ""
\ No newline at end of file
diff --git a/python/tools/memory_delete.py b/python/tools/memory_delete.py
new file mode 100644
index 000000000..b692c5430
--- /dev/null
+++ b/python/tools/memory_delete.py
@@ -0,0 +1,11 @@
+from python.helpers.memory import Memory
+from python.helpers.tool import Tool, Response
+
+class MemoryForget(Tool):
+
+ async def execute(self, ids=[], **kwargs):
+ db = await Memory.get(self.agent)
+ dels = await db.delete_documents_by_ids(ids=ids)
+
+ result = self.agent.read_prompt("fw.memories_deleted.md", memory_count=dels)
+ return Response(message=result, break_loop=False)
\ No newline at end of file
diff --git a/python/tools/memory_forget.py b/python/tools/memory_forget.py
new file mode 100644
index 000000000..dd97a95ec
--- /dev/null
+++ b/python/tools/memory_forget.py
@@ -0,0 +1,13 @@
+from python.helpers.memory import Memory
+from python.helpers.tool import Tool, Response
+
+DEFAULT_THRESHOLD = 0.75
+
+class MemoryForget(Tool):
+
+ async def execute(self, query="", threshold=DEFAULT_THRESHOLD, filter="", **kwargs):
+ db = await Memory.get(self.agent)
+ dels = await db.delete_documents_by_query(query=query, threshold=threshold, filter=filter)
+
+ result = self.agent.read_prompt("fw.memories_deleted.md", memory_count=len(dels))
+ return Response(message=result, break_loop=False)
\ No newline at end of file
diff --git a/python/tools/memory_load.py b/python/tools/memory_load.py
new file mode 100644
index 000000000..de08459fd
--- /dev/null
+++ b/python/tools/memory_load.py
@@ -0,0 +1,19 @@
+from python.helpers.memory import Memory
+from python.helpers.tool import Tool, Response
+
+DEFAULT_THRESHOLD = 0.6
+DEFAULT_LIMIT = 10
+
+class MemoryLoad(Tool):
+
+ async def execute(self, query="", threshold=DEFAULT_THRESHOLD, limit=DEFAULT_LIMIT, filter="", **kwargs):
+ db = await Memory.get(self.agent)
+ docs = await db.search_similarity_threshold(query=query, limit=limit, threshold=threshold, filter=filter)
+
+ if len(docs) == 0:
+ result = self.agent.read_prompt("fw.memories_not_found.md", query=query)
+ else:
+ text = "\n\n".join(Memory.format_docs_plain(docs))
+ result = str(text)
+
+ return Response(message=result, break_loop=False)
\ No newline at end of file
diff --git a/python/tools/memory_save.py b/python/tools/memory_save.py
new file mode 100644
index 000000000..be2b86cca
--- /dev/null
+++ b/python/tools/memory_save.py
@@ -0,0 +1,20 @@
+from python.helpers.memory import Memory
+from python.helpers.tool import Tool, Response
+
+DEFAULT_THRESHOLD = 0.5
+DEFAULT_LIMIT = 5
+
+class MemorySave(Tool):
+
+ async def execute(self, text="", area="", **kwargs):
+
+ if not area:
+ area = Memory.Area.MAIN.value
+
+ metadata = {"area": area, **kwargs}
+
+ db = await Memory.get(self.agent)
+ id = db.insert_text(text, metadata)
+
+ result = self.agent.read_prompt("fw.memory_saved.md", memory_id=id)
+ return Response(message=result, break_loop=False)
diff --git a/python/tools/memory_tool.py b/python/tools/memory_tool.py
deleted file mode 100644
index 11cfdf005..000000000
--- a/python/tools/memory_tool.py
+++ /dev/null
@@ -1,75 +0,0 @@
-import re
-from agent import Agent
-from python.helpers.vector_db import get_or_create_db
-import os
-from python.helpers.tool import Tool, Response
-from python.helpers.print_style import PrintStyle
-from python.helpers.errors import handle_error
-from python.helpers import files
-
-class Memory(Tool):
- async def execute(self,**kwargs):
- result=""
-
- area = kwargs.get("area", "manual") # when called by agent, it will always be manual
-
- try:
- if "query" in kwargs:
- threshold = float(kwargs.get("threshold", 0.1))
- count = int(kwargs.get("count", 5))
- result = search(self.agent, kwargs["query"], count, threshold)
- elif "memorize" in kwargs:
- result = save(self.agent, kwargs["memorize"])
- elif "forget" in kwargs:
- result = forget(self.agent, kwargs["forget"])
- # elif "delete" in kwargs
- result = delete(self.agent, kwargs["delete"])
- except Exception as e:
- handle_error(e)
- # hint about embedding change with existing database
- PrintStyle.hint("If you changed your embedding model, you will need to remove contents of /memory directory.")
- self.agent.context.log.log(type="hint", content="If you changed your embedding model, you will need to remove contents of /memory directory.")
- raise
-
- # result = process_query(self.agent, self.args["memory"],self.args["action"], result_count=self.agent.config.auto_memory_count)
- return Response(message=result, break_loop=False)
-
-def search(agent:Agent, query:str, count:int=5, threshold:float=0.1):
- db = get_db(agent)
- # docs = db.search_similarity(query,count) # type: ignore
- docs = db.search_similarity_threshold(query,count,threshold) # type: ignore
- if len(docs)==0: return agent.read_prompt("fw.memories_not_found.md", query=query)
- else: return str(docs)
-
-def save(agent:Agent, text:str):
- db = get_db(agent)
- id = db.insert_text(text) # type: ignore
- return agent.read_prompt("fw.memory_saved.md", memory_id=id)
-
-def delete(agent:Agent, ids_str:str):
- db = get_db(agent)
- ids = extract_guids(ids_str)
- deleted = db.delete_documents_by_ids(ids) # type: ignore
- return agent.read_prompt("fw.memories_deleted.md", memory_count=deleted)
-
-def forget(agent:Agent, query:str):
- db = get_db(agent)
- deleted = db.delete_documents_by_query(query) # type: ignore
- return agent.read_prompt("fw.memories_deleted.md", memory_count=deleted)
-
-def get_db(agent: Agent):
- mem_dir = files.get_abs_path("memory", agent.config.memory_subdir or "default")
- kn_dirs = [files.get_abs_path("knowledge", d) for d in agent.config.knowledge_subdirs or []]
-
- db = get_or_create_db(
- agent.context.log,
- embeddings_model=agent.config.embeddings_model,
- in_memory=False,
- memory_dir=mem_dir,
- knowledge_dirs=kn_dirs)
-
- return db
-
-def extract_guids(text):
- pattern = r'\b[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-5][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}\b'
- return re.findall(pattern, text)
\ No newline at end of file
diff --git a/python/tools/memory_tool.py.txt b/python/tools/memory_tool.py.txt
new file mode 100644
index 000000000..63a7637c1
--- /dev/null
+++ b/python/tools/memory_tool.py.txt
@@ -0,0 +1,92 @@
+import re
+from agent import Agent
+from python.helpers.vector_db import get_or_create_db, Area
+import os
+from python.helpers.tool import Tool, Response
+from python.helpers.print_style import PrintStyle
+from python.helpers.errors import handle_error
+from python.helpers import files
+
+DEFAULT_THRESHOLD = 0.5
+
+
+class Memory(Tool):
+
+ async def execute(self, **kwargs):
+ result = ""
+
+ try:
+ if "query" in kwargs:
+ threshold = float(kwargs.get("threshold", DEFAULT_THRESHOLD))
+ count = int(kwargs.get("limit", 5))
+ result = search(self.agent, kwargs["query"], count, threshold)
+ elif "memorize" in kwargs:
+ meta = {"area": Area.MAIN.value}
+ result = save(self.agent, kwargs["memorize"])
+ elif "forget" in kwargs:
+ result = forget(self.agent, kwargs["forget"])
+ # elif "delete" in kwargs
+ result = delete(self.agent, kwargs["delete"])
+ except Exception as e:
+ handle_error(e)
+ # hint about embedding change with existing database
+ PrintStyle.hint(
+ "If you changed your embedding model, you will need to remove contents of /memory directory."
+ )
+ self.agent.context.log.log(
+ type="hint",
+ content="If you changed your embedding model, you will need to remove contents of /memory directory.",
+ )
+ raise
+
+ # result = process_query(self.agent, self.args["memory"],self.args["action"], result_count=self.agent.config.auto_memory_count)
+ return Response(message=result, break_loop=False)
+
+
+def search(
+ agent: Agent, query: str, count: int = 5, threshold: float = DEFAULT_THRESHOLD
+):
+ db = get_db(agent)
+ # docs = db.search_similarity(query,count) # type: ignore
+ docs = db.search_similarity_threshold(query=query, limit=count, threshold=threshold) # type: ignore
+ if len(docs) == 0:
+ return agent.read_prompt("fw.memories_not_found.md", query=query)
+ else:
+ return str(docs)
+
+
+def save(agent: Agent, text: str, metadata: dict = {}):
+ db = get_db(agent)
+ id = db.insert_text(text, metadata) # type: ignore
+ return agent.read_prompt("fw.memory_saved.md", memory_id=id)
+
+
+def delete(agent: Agent, ids_str: str):
+ db = get_db(agent)
+ ids = extract_guids(ids_str)
+ deleted = db.delete_documents_by_ids(ids) # type: ignore
+ return agent.read_prompt("fw.memories_deleted.md", memory_count=deleted)
+
+
+def forget(agent: Agent, query: str):
+ db = get_db(agent)
+ deleted = db.delete_documents_by_query(query) # type: ignore
+ return agent.read_prompt("fw.memories_deleted.md", memory_count=deleted)
+
+
+def get_db(agent: Agent):
+ mem_dir = files.get_abs_path("memory", agent.config.memory_subdir or "default")
+ kn_dirs = [
+ files.get_abs_path("knowledge", d) for d in agent.config.knowledge_subdirs or []
+ ]
+
+ db = get_or_create_db(
+ agent=agent,
+ )
+
+ return db
+
+
+def extract_guids(text):
+ pattern = r"\b[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-5][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}\b"
+ return re.findall(pattern, text)
diff --git a/python/tools/unknown.py b/python/tools/unknown.py
index 3f52c9af1..96c7f856f 100644
--- a/python/tools/unknown.py
+++ b/python/tools/unknown.py
@@ -1,10 +1,15 @@
from python.helpers.tool import Tool, Response
+from python.extensions.message_loop_prompts._10_tool_instructions import (
+ concat_tool_prompts,
+)
+
class Unknown(Tool):
async def execute(self, **kwargs):
+ tools = concat_tool_prompts(self.agent)
return Response(
- message=self.agent.read_prompt("fw.tool_not_found.md",
- tool_name=self.name,
- tools_prompt=self.agent.read_prompt("agent.tools.md")),
- break_loop=False)
-
+ message=self.agent.read_prompt(
+ "fw.tool_not_found.md", tool_name=self.name, tools_prompt=tools
+ ),
+ break_loop=False,
+ )
diff --git a/requirements.txt b/requirements.txt
index b32c05443..08ef38c7a 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -12,7 +12,6 @@ langchain_mistralai==0.1.8
webcolors==24.6.0
sentence-transformers==3.0.1
docker==7.1.0
-pandas==2.2.3
paramiko==3.4.0
duckduckgo_search==6.1.12
inputimeout==1.0.4
diff --git a/webui/index.css b/webui/index.css
index 531f51da8..eb3c83f4b 100644
--- a/webui/index.css
+++ b/webui/index.css
@@ -15,7 +15,7 @@
--color-input-focus-dark: #1b1b1b;
/* Light mode */
- --color-background-light: #F5F7FA;
+ --color-background-light: #e8e9e9;
--color-text-light: #333333;
--color-primary-light: #273b4d;
--color-secondary-light: #e8eaf6;
@@ -312,6 +312,7 @@ h4 {
.message-user {
background-color: #4a4a4a;
border-bottom-right-radius: var(--spacing-xs);
+ text-align: end;
}
.message-ai {
@@ -345,15 +346,15 @@ h4 {
/* Update message types for dark mode */
.message-default { background-color: #1A242F; color: #E0E0E0; }
-.message-agent { background-color: #2C3E50; color: #E0E0E0; }
-.message-agent-response { background-color: #002b54; color: #E0E0E0; }
-.message-agent-delegation { background-color: #00695C; color: #E0E0E0; }
-.message-tool { background-color: #5D4C6A; color: #E0E0E0; }
-.message-code-exe { background-color: #3a147c; color: #E0E0E0; }
+.message-agent { background-color: #34506b; color: #E0E0E0; }
+.message-agent-response { background-color: #1f3c1e; color: #E0E0E0; }
+.message-agent-delegation { background-color: #12685e; color: #E0E0E0; }
+.message-tool { background-color: #2a4170; color: #E0E0E0; }
+.message-code-exe { background-color: #4b3a69; color: #E0E0E0; }
.message-info { background-color: var(--color-panel); color: #E0E0E0; }
.message-util { background-color: #23211a; color: #E0E0E0; display:none }
-.message-warning { background-color: #c2771b; color: #E0E0E0; }
-.message-error { background-color: #ab1313; color: #E0E0E0; }
+.message-warning { background-color: #bc8036; color: #E0E0E0; }
+.message-error { background-color: #af2222; color: #E0E0E0; }
/* Agent and AI Info */
.agent-start {
@@ -615,18 +616,18 @@ input:checked + .slider:before {
--color-input-focus: var(--color-input-focus-light);
}
-.light-mode .message-default { background-color: #E3F2FD; color: #1A242F; }
-.light-mode .message-agent { background-color: #e9ebf3; color: #2C3E50; }
-.light-mode .message-agent-response { background-color: #E1E6F2; color: #002b54; }
-.light-mode .message-agent-delegation { background-color: #E0F2F1; color: #00695C; }
-.light-mode .message-tool { background-color: #EDE7F6; color: #5D4C6A; }
-.light-mode .message-code-exe { background-color: #FCE4EC; color: #3a147c; }
-.light-mode .message-info { background-color: #E8EAF6; color: #2C3E50; }
-.light-mode .message-util { background-color: #e8eaf6d6; color: #353c43; }
-.light-mode .message-warning { background-color: #FFF3E0; color: #c2771b; }
-.light-mode .message-error { background-color: #FFEBEE; color: #ab1313; }
+.light-mode .message-default { background-color: #ffffff; color: #1A242F; }
+.light-mode .message-agent { background-color: #ffffff; color: #356ca3; }
+.light-mode .message-agent-response { background-color: #ffffff; color: #188216; }
+.light-mode .message-agent-delegation { background-color: #ffffff; color: #12685e; }
+.light-mode .message-tool { background-color: #ffffff; color: #1c3c88; }
+.light-mode .message-code-exe { background-color: #ffffff; color: #6c43b0; }
+.light-mode .message-info { background-color: #ffffff; color: #3f3f3f; }
+.light-mode .message-util { background-color: #ffffff; color: #5b5540; }
+.light-mode .message-warning { background-color: #ffffff; color: #8f4800; }
+.light-mode .message-error { background-color: #ffffff; color: #8f1010; }
.light-mode .message-user {
- background-color: #eaeaea;
+ background-color: #ffffff;
color: #4e4e4e;
}
diff --git a/webui/index.html b/webui/index.html
index 09b7a118f..016afb559 100644
--- a/webui/index.html
+++ b/webui/index.html
@@ -70,7 +70,7 @@
<li x-data="{ autoScroll: true }">
<span>Autoscroll</span>
<label class="switch">
- <input type="checkbox" x-model="autoScroll"
+ <input id="auto-scroll-switch" type="checkbox" x-model="autoScroll"
x-effect="window.safeCall('toggleAutoScroll',autoScroll)">
<span class="slider"></span>
</label>
@@ -123,14 +123,14 @@
<!--Chat-->
<div id="chat-history">
</div>
- <div id="progress-bar-box">
- <h4 id="progress-bar-h"><span id="progress-bar-i">|></span><span id="progress-bar"></span></h4>
- </div>
<div id="toast" class="toast">
<div class="toast__message"></div>
<button class="toast__copy">Copy</button>
<button class="toast__close">Close</button>
</div>
+ <div id="progress-bar-box">
+ <h4 id="progress-bar-h"><span id="progress-bar-i">|></span><span id="progress-bar"></span></h4>
+ </div>
<div id="input-section" x-data="{ paused: false }">
<textarea id="chat-input" placeholder="Type your message here..." rows="1"></textarea>
<button class="chat-button" id="send-button" aria-label="Send message">
diff --git a/webui/index.js b/webui/index.js
index 7afed4769..56ecdd83f 100644
--- a/webui/index.js
+++ b/webui/index.js
@@ -11,6 +11,7 @@ const statusSection = document.getElementById('status-section');
const chatsSection = document.getElementById('chats-section');
const scrollbarThumb = document.querySelector('#chat-history::-webkit-scrollbar-thumb');
const progressBar = document.getElementById('progress-bar');
+const autoScrollSwitch = document.getElementById('auto-scroll-switch');
@@ -395,6 +396,23 @@ function toast(text, type = 'info') {
}, 10000);
}
+function scrollChanged(isAtBottom) {
+ const inputAS = Alpine.$data(autoScrollSwitch);
+ inputAS.autoScroll = isAtBottom
+ // autoScrollSwitch.checked = isAtBottom
+ console.log(isAtBottom)
+}
+
+chatHistory.addEventListener('scroll', function () {
+ // const toleranceEm = 1; // Tolerance in em units
+ // const tolerancePx = toleranceEm * parseFloat(getComputedStyle(document.documentElement).fontSize); // Convert em to pixels
+ const tolerancePx = 50;
+ const chatHistory = document.getElementById('chat-history');
+ const isAtBottom = (chatHistory.scrollHeight - chatHistory.scrollTop) <= (chatHistory.clientHeight + tolerancePx);
+
+ scrollChanged(isAtBottom);
+});
+
chatInput.addEventListener('input', adjustTextareaHeight);
setInterval(poll, 250);