Migrate to PydanticAI for LLM and summary agents

.github/workflows/python-app.yml

@@ -14,7 +14,7 @@ jobs:
   build:
     runs-on: ubuntu-latest
     container:
-      image: python:3.10-slim
+      image: python:3.13-slim
 
     steps:
     - uses: actions/checkout@v4
@@ -23,6 +23,7 @@ jobs:
         python -m pip install --upgrade pip
         pip install flake8 pytest
         if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+        if [ -f requirements-dev.txt ]; then pip install -r requirements-dev.txt; fi
         # Install the package in development mode
         pip install -e .
     - name: Lint with flake8

CLAUDE.md

@@ -4,7 +4,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Project Overview
 
-InnieMe is a Discord bot that provides AI-powered Q&A capabilities using document knowledge bases. The bot scans and vectorizes documents from specified directories, connects to Discord channels, and responds to user mentions with context-aware responses using OpenAI's GPT models.
+InnieMe is a Discord bot that provides AI-powered Q&A capabilities using document knowledge bases. The bot scans and vectorizes documents from specified directories, connects to Discord channels, and responds to user mentions with context-aware responses using configurable LLM providers via PydanticAI.
 
 ## Common Development Commands
 
@@ -16,7 +16,7 @@ pip install -r requirements-dev.txt
 
 # Create configuration from example
 cp config.example.yaml config.yaml
-# Edit config.yaml with your Discord token, OpenAI API key, and channel settings
+# Edit config.yaml with your Discord token, API keys, LLM model, and channel settings
 ```
 
 ### Running the Bot
@@ -61,9 +61,9 @@ The application follows a modular architecture with clear separation of concerns
 1. **DiscordBot** (`src/innieme/discord_bot.py`): Main bot interface that handles Discord events, commands, and message routing
 2. **Innie** (`src/innieme/innie.py`): Container class that manages multiple topics and their configurations
 3. **Topic** (`src/innieme/innie.py`): Represents a single topic with its own document store, channels, and conversation engine
-4. **ConversationEngine** (`src/innieme/conversation_engine.py`): Handles query processing and response generation using OpenAI
+4. **ConversationEngine** (`src/innieme/conversation_engine.py`): Handles query processing and response generation using a PydanticAI `Agent`
 5. **DocumentProcessor** (`src/innieme/document_processor.py`): Manages document scanning, vectorization, and similarity search
-6. **KnowledgeManager** (`src/innieme/knowledge_manager.py`): Handles conversation summarization and knowledge base storage
+6. **KnowledgeManager** (`src/innieme/knowledge_manager.py`): Handles conversation summarization (via a PydanticAI `Agent` with structured `SummaryOutput`) and knowledge base storage
 
 ### Factory Pattern Components
 
@@ -78,6 +78,12 @@ The bot uses YAML configuration (`config.yaml`) with the following structure:
 - Each topic has its own role/system prompt, document directory, and Discord channels
 - Configuration is loaded via `DiscordBotConfig` class
 
+Key top-level config fields:
+- `embedding_model`: `"openai"`, `"huggingface"`, or `"fake"`
+- `embeddings_api_key`: API key for the embedding model (required when `embedding_model` is `"openai"`)
+- `llm_model`: PydanticAI model string, e.g. `"openai:gpt-4o"` or `"anthropic:claude-sonnet-4-6"`
+- `llm_api_key`: API key for the LLM provider
+
 ### Bot Behavior
 
 - Bot responds when mentioned in Discord channels
@@ -99,8 +105,9 @@ The bot uses YAML configuration (`config.yaml`) with the following structure:
 - Document search provides context for LLM responses
 
 ### Response Generation
-- Uses OpenAI GPT-3.5-turbo model by default
-- Combines document context with conversation history
+- Uses PydanticAI `Agent` with a configurable LLM (default: `openai:gpt-3.5-turbo`)
+- LLM provider and model are set via `llm_model` in `config.yaml` (e.g. `"openai:gpt-4o"`, `"anthropic:claude-sonnet-4-6"`)
+- Combines document context with conversation history via `ConversationDependencies`
 - Handles responses longer than Discord's 2000 character limit by sending as files
 
 ### Error Handling

config.example.yaml

@@ -8,16 +8,22 @@
 # 6. Paste your token below (keep it secret!)
 discord_token: discord_bot_token
 
-# OpenAI API (for embeddings)
-# To get your OpenAI API key:
-# 1. Go to OpenAI's website (https://platform.openai.com)
-# 2. Sign in or create an account
-# 3. Click on your profile icon and select "View API keys"
-# 4. Click "Create new secret key"
-# 5. Copy your API key (you won't be able to see it again!)
-# 6. Paste your key below (keep it secret!)
-openai_api_key: openai_api_key
-embedding_model: "openai" # Options: "openai", "huggingface"
+# Embedding model to use for document vectorisation
+# Options: "openai", "huggingface", "fake"
+embedding_model: "openai"
+
+# API key for the embedding model (only required when embedding_model is "openai")
+embeddings_api_key: openai_api_key
+
+# LLM to use for conversation and summarisation
+# Format: "<provider>:<model>", e.g. "openai:gpt-4o", "anthropic:claude-sonnet-4-6"
+llm_model: "openai:gpt-3.5-turbo"
+
+# API key for the LLM provider
+# For OpenAI: https://platform.openai.com/api-keys
+# For Anthropic: https://console.anthropic.com/settings/keys
+llm_api_key: llm_api_key
+
 outies:
 # To get your Discord admin user ID:
 # 1. Go to Discord and go to User Settings (gear icon)
@@ -54,4 +60,4 @@ outies:
         docs_dir: "./data/documents2"
         channels:
           - guild_id: discord_server_id
-            channel_id: discord_channel_id
+            channel_id: discord_channel_id

pyproject.toml

@@ -6,15 +6,33 @@ build-backend = "setuptools.build_meta"
 name = "innieme"
 version = "0.1.0"
 description = "Your bot description"
-requires-python = ">=3.9"
+requires-python = ">=3.13"
 dependencies = [
-    # Your dependencies here
+    "discord.py",
+    "python-dotenv",
+    "chromadb",
+    "langchain>=0.1.0,<1.0.0",
+    "langchain-community",
+    "langchain-chroma",
+    "langchain-huggingface",
+    "langchain-openai",
+    "pypdf",
+    "python-docx",
+    "faiss-cpu",
+    "numpy",
+    "openai>=1.0.0",
+    "pydantic-ai>=1.0.0",
+    "audioop-lts==0.2.1; python_version >= '3.13'",
 ]
 
 [project.optional-dependencies]
 dev = [
     "pytest",
-    # Other development dependencies
+    "pytest-asyncio",
+    "pytest-cov",
+    "black",
+    "flake8",
+    "isort",
 ]
 
 [project.scripts]