Dev/add chroma (#5)

* chroma added but session state lost never luck

* working chroma plus async comms with persistent history

* store chroma db

* refac

* refactor

* add ruff

* update pipeline

* fix ruff

Author: JakubS-Kentico

.github/copilot-instructions.md

@@ -0,0 +1,19 @@
+You are "Synapse", an AI Senior Tech Lead specialized in guiding the development of AI agents. Your role is to help users (developers) write robust, efficient, and maintainable code for AI agent systems.
+
+**Core Focus:**
+
+* **Agent Architecture:** Advise on agent design patterns (e.g., ReAct, Plan-and-Execute), state management, and modularity.
+* **LLM Interaction:** Guide effective prompt engineering, context window management, token optimization, and handling LLM errors/limitations.
+* **Tooling & Integration:** Promote reliable and secure integration of tools/APIs, including robust error handling.
+* **Code Quality:** Enforce clean, testable, and maintainable code principles within the agent's implementation.
+* **Observability & Debugging:** Recommend strategies for logging agent behavior, decisions, and tool usage to facilitate debugging and evaluation.
+* **Testing:** Suggest practical approaches for testing agent reliability and task success.
+
+**Interaction Style:**
+
+* Provide concise, actionable feedback on code and designs.
+* Explain the reasoning behind your recommendations, focusing on AI agent best practices.
+* Ask clarifying questions to understand context.
+* Be pragmatic, balancing ideal solutions with practical constraints.
+
+**Your Goal:** Mentor the developer to build high-quality, effective AI agents.

.github/workflows/build-test.yaml

@@ -18,17 +18,42 @@ jobs:
     - name: Set up Python
       uses: actions/setup-python@v5
       with:
-        python-version: '3.11' # Or your desired Python version
+        # Use the Python version specified in pyproject.toml
+        # Make sure this matches your requires-python range
+        python-version: '3.12'
+
+    - name: Install Poetry
+      uses: snok/install-poetry@v1
+      with:
+        virtualenvs-create: true # Recommended to keep envs isolated
+        virtualenvs-in-project: true # Optional: Keep venv in project dir for caching
+        installer-parallel: true
+
+    # Optional: Cache Poetry virtualenv
+    - name: Load cached venv
+      id: cached-poetry-dependencies
+      uses: actions/cache@v4
+      with:
+        path: .venv
+        key: venv-${{ runner.os }}-${{ steps.setup-python.outputs.python-version }}-${{ hashFiles('**/poetry.lock') }}
 
     - name: Install dependencies
-      run: |
-        python -m pip install --upgrade pip
-        pip install -r requirements.txt
+      if: steps.cached-poetry-dependencies.outputs.cache-hit != 'true'
+      run: poetry install --no-interaction --no-root # Install dependencies only
+
+    # Install project and dependencies if cache missed or lock file changed
+    - name: Install project and dependencies
+      run: poetry install --no-interaction
 
     - name: Lint with Ruff
       run: |
-        ruff check .
+        poetry run ruff check .
 
     - name: Check formatting with Ruff
       run: |
-        ruff format --check .
+        poetry run ruff format --check .
+
+    # Add other steps like running tests using poetry run pytest (if you have tests)
+    # - name: Run tests
+    #   run: |
+    #     poetry run pytest

.gitignore

@@ -176,3 +176,9 @@ config.json
 repomix-output.*
 memory.json
 .flask_session_files
+
+# Cachelib history cache
+.chat_history_cache/
+
+# Flask Session files (if Flask-Session is kept)
+.flask_session_files/

bot_config/config.json.sample

@@ -5,10 +5,13 @@
     "D:/another/project/path",
     "/path/on/linux/or/mac"
   ],
-  "enable_memory_server": true,
+  "enable_memory_server": false,
+  "enable_chroma_server": true,
+  "chroma_path": "./chroma_db",
+  "chroma_collection_name": "chat_history",
   "default_gemini_model": "gemini-1.5-flash",
   "generation_gemini_model": "gemini-1.5-flash",
-  "summarization_gemini_model": "gemini-1.5-flash"
+  "summarization_gemini_model": "gemini-1.5-flash",
   "max_debug_log_size": 1500,
   "log_preview_len": 250
 }

bot_config/system_instruction.md

@@ -2,8 +2,8 @@ You are just a rather very intelligent system with access to tools for interacti
 
 - You can use these tools sequentially within a single user turn to fulfill complex requests. If a request requires multiple steps (like reading one file to find the name of another file to read), make the necessary function calls one after another. When your are done querying tools, answer the user with a comprehensive response.
 
-- If you dont understand query, refer to the 'session_summary' field to understand the ongoing conversation. Use the 'search_nodes' function to search for the field.
+- If you dont understand query, refer to the 'chat_history' chroma collection to understand the ongoing conversation. Use the 'chroma_query_documents' function to search.
 
-- After query, update the 'session_summary' field in the knowledge graph. Keep last 10 topics. Keep 100 characters per topic.
+- After query, update the 'chat_history' to keep records. Keep the last 10 topics. Keep 100 characters per topic.
 
 - After done using all the tools required, reply to user.