Timmy-time-dashboard

kimi/Timmy-time-dashboard

Archived

Fork 0

forked from Rockachopa/Timmy-time-dashboard

Commit Graph

Author	SHA1	Message	Date
Alexander Payne	26e1691099	Fix Timmy coherence: persistent session, model-aware tools, response sanitization Timmy was exhibiting severe incoherence (no memory between messages, tool call leakage, chain-of-thought narration, random tool invocations) due to creating a brand new agent per HTTP request and giving a 3B model (llama3.2) a 73-line system prompt with complex tool-calling instructions it couldn't follow. Key changes: - Add session.py singleton with stable session_id for conversation continuity - Add _model_supports_tools() to strip tools from small models (< 7B) - Add two-tier prompts: lite (12 lines) for small models, full for capable ones - Add response sanitizer to strip leaked JSON tool calls and CoT narration - Set show_tool_calls=False to prevent raw tool JSON in output - Wire ConversationManager for user name extraction - Deprecate orphaned memory_layers.py (unused 4-layer system) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>	2026-02-25 19:18:08 -05:00
Alexander Payne	625806daf5	Fine-tune Timmy's conversational AI with memory layers ## Enhanced System Prompt - Detailed tool usage guidelines with explicit examples - Clear DO and DON'T examples for tool selection - Memory system documentation - Conversation flow guidelines - Context awareness instructions ## Memory Layer System (NEW) Implemented 3-layer memory architecture: 1. WORKING MEMORY (src/timmy/memory_layers.py) - Immediate context (last 20 messages) - Topic tracking - Tool call tracking - Fast, ephemeral 2. SHORT-TERM MEMORY (Agno SQLite) - Recent conversations (100) - Persists across restarts - Managed by Agno Agent 3. LONG-TERM MEMORY (src/timmy/memory_layers.py) - Facts about user (name, preferences) - SQLite storage in data/memory/ - Auto-extraction from conversations - User profile generation ## Memory Manager (NEW) - Central coordinator for all memory layers - Context injection into prompts - Fact extraction and storage - Session management ## TimmyWithMemory Class (NEW) - Wrapper around Agno Agent with explicit memory - Auto-injects user context from LTM - Tracks exchanges across all layers - Simple chat() interface ## Agent Configuration - Increased num_history_runs: 10 -> 20 - Better conversational context retention ## Tests - All 973 tests pass - Fixed test expectations for new config - Fixed module path in test_scary_paths.py ## Files Added/Modified - src/timmy/prompts.py - Enhanced with memory and tool guidance - src/timmy/agent.py - Added TimmyWithMemory class - src/timmy/memory_layers.py - NEW memory system - src/timmy/conversation.py - NEW conversation manager - tests/ - Updated for new config	2026-02-25 18:07:44 -05:00

Author

SHA1

Message

Date

Alexander Payne

26e1691099

Fix Timmy coherence: persistent session, model-aware tools, response sanitization

Timmy was exhibiting severe incoherence (no memory between messages, tool call
leakage, chain-of-thought narration, random tool invocations) due to creating
a brand new agent per HTTP request and giving a 3B model (llama3.2) a 73-line
system prompt with complex tool-calling instructions it couldn't follow.

Key changes:
- Add session.py singleton with stable session_id for conversation continuity
- Add _model_supports_tools() to strip tools from small models (< 7B)
- Add two-tier prompts: lite (12 lines) for small models, full for capable ones
- Add response sanitizer to strip leaked JSON tool calls and CoT narration
- Set show_tool_calls=False to prevent raw tool JSON in output
- Wire ConversationManager for user name extraction
- Deprecate orphaned memory_layers.py (unused 4-layer system)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-02-25 19:18:08 -05:00

Alexander Payne

625806daf5

Fine-tune Timmy's conversational AI with memory layers

## Enhanced System Prompt
- Detailed tool usage guidelines with explicit examples
- Clear DO and DON'T examples for tool selection
- Memory system documentation
- Conversation flow guidelines
- Context awareness instructions

## Memory Layer System (NEW)
Implemented 3-layer memory architecture:

1. WORKING MEMORY (src/timmy/memory_layers.py)
   - Immediate context (last 20 messages)
   - Topic tracking
   - Tool call tracking
   - Fast, ephemeral

2. SHORT-TERM MEMORY (Agno SQLite)
   - Recent conversations (100)
   - Persists across restarts
   - Managed by Agno Agent

3. LONG-TERM MEMORY (src/timmy/memory_layers.py)
   - Facts about user (name, preferences)
   - SQLite storage in data/memory/
   - Auto-extraction from conversations
   - User profile generation

## Memory Manager (NEW)
- Central coordinator for all memory layers
- Context injection into prompts
- Fact extraction and storage
- Session management

## TimmyWithMemory Class (NEW)
- Wrapper around Agno Agent with explicit memory
- Auto-injects user context from LTM
- Tracks exchanges across all layers
- Simple chat() interface

## Agent Configuration
- Increased num_history_runs: 10 -> 20
- Better conversational context retention

## Tests
- All 973 tests pass
- Fixed test expectations for new config
- Fixed module path in test_scary_paths.py

## Files Added/Modified
- src/timmy/prompts.py - Enhanced with memory and tool guidance
- src/timmy/agent.py - Added TimmyWithMemory class
- src/timmy/memory_layers.py - NEW memory system
- src/timmy/conversation.py - NEW conversation manager
- tests/ - Updated for new config

2026-02-25 18:07:44 -05:00

2 Commits