Support Ollama and unify OpenAI client

- Add LLM_PROVIDER to .env.example and include Ollama config/examples
- Update README, SETUP_CHECKLIST, and copilot instructions to document provider choice and Ollama setup
- Remove azure_openai_client.py and add unified openai_client.py (UnifiedOpenAIClient) that supports both Azure OpenAI and Ollama
- Update brain_core/config.py to validate LLM_PROVIDER, load provider-specific settings, and return the UnifiedOpenAIClient

Author: madebygps

.env.example

@@ -1,18 +1,39 @@
-# Diary Configuration
+# Second Brain Configuration
 # Copy this file to .env and fill in your paths
+#
+# LLM Provider Options:
+#   - azure: Cloud-based, requires API key, costs money, very capable
+#   - ollama: Local, free, private, no internet needed (requires installation)
+#
+# Quick Start with Ollama (Free & Local):
+#   1. Install: https://ollama.com
+#   2. Run: ollama pull llama3.1
+#   3. Set: LLM_PROVIDER=ollama
+#
+# Quick Start with Azure OpenAI (Cloud):
+#   1. Get API key from Azure Portal
+#   2. Set: LLM_PROVIDER=azure
+#   3. Fill in AZURE_OPENAI_* settings below
 
 # Required: Path to your Obsidian vault (or any markdown directory)
 DIARY_PATH=/path/to/obsidian/vault
 
 # Required: Path to planner directory for extracted todos
 PLANNER_PATH=/path/to/planner
 
-# Required: Azure OpenAI configuration
+# Required: LLM Provider ("azure" or "ollama")
+LLM_PROVIDER=azure
+
+# Azure OpenAI Configuration (required if LLM_PROVIDER=azure)
 AZURE_OPENAI_API_KEY=your-api-key-here
 AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
 AZURE_OPENAI_DEPLOYMENT=gpt-4o
 AZURE_OPENAI_API_VERSION=2024-02-15-preview
 
+# Ollama Configuration (required if LLM_PROVIDER=ollama)
+# OLLAMA_BASE_URL=http://localhost:11434
+# OLLAMA_MODEL=llama3.1
+
 # Optional: Cost Tracking Configuration
 BRAIN_COST_DB_PATH=~/.brain/costs.db
 

.github/copilot-instructions.md

@@ -9,7 +9,7 @@ AI-powered journaling system with markdown entries, semantic backlinks, and LLM-
 - Package manager: `uv` (ALWAYS use `uv`, never `pip`)
 - Testing: pytest with 57 tests, 47% coverage
 - CLI: typer (add_completion=False) + rich
-- LLM: Azure OpenAI (required)
+- LLM: Azure OpenAI or Ollama (configurable via LLM_PROVIDER)
 
 ## Architecture
 
@@ -22,7 +22,7 @@ Core functionality with comprehensive type hints and dataclasses:
 - `constants.py` - Shared constants (100% coverage)
 - `entry_manager.py` - Entry I/O and parsing (93% coverage)
 - `llm_client.py` - Abstract LLM interface (73% coverage)
-- `azure_openai_client.py` - Azure OpenAI client (40% coverage)
+- `openai_client.py` - Unified OpenAI client for both Azure and Ollama (0% coverage)
 - `llm_analysis.py` - Atomic LLM operations: entities, backlinks, tags (17% coverage)
 - `report_generator.py` - Report orchestration (18% coverage)
 - `template_generator.py` - AI prompt generation (42% coverage)
@@ -63,12 +63,20 @@ uv run pytest tests/ --cov        # With coverage
 - `DIARY_PATH` - Path to Obsidian vault or markdown directory (for reflection entries)
 - `PLANNER_PATH` - Path to directory for daily plan files (separate from diary)
 
-**Azure OpenAI (required for all LLM features):**
+**LLM Provider (choose one):**
+- `LLM_PROVIDER` - Set to "azure" or "ollama" (default: azure)
+
+**Azure OpenAI (cloud-based):**
 - `AZURE_OPENAI_API_KEY` - API key
 - `AZURE_OPENAI_ENDPOINT` - Service endpoint
 - `AZURE_OPENAI_DEPLOYMENT` - Model deployment name (e.g., gpt-4o)
 - `AZURE_OPENAI_API_VERSION` - Default: 2024-02-15-preview
-- ✅ **Full functionality:** Diary prompts, task extraction, semantic analysis, backlinks, tags, reports
+- ✅ **Full functionality:** All features, cost tracking enabled
+
+**Ollama (local, free):**
+- `OLLAMA_BASE_URL` - API URL (default: http://localhost:11434)
+- `OLLAMA_MODEL` - Model name (default: llama3.1)
+- ✅ **Full functionality:** All features, no cost tracking (local is free!)
 
 ## Entry Structure
 
@@ -176,16 +184,21 @@ llm_client.py → azure_openai_client.py
 ## Recent Refactoring
 
 **Architecture Simplification:**
-- Removed Ollama client (195 lines) - Azure OpenAI only
 - Removed scheduler/daemon system (312 lines, 0% coverage)
 - Removed todos command (redundant with plan command)
 - Removed `generate_planning_prompts()` - plans no longer have prompts
 - Removed notes search functionality (238 lines) - focused on journaling/planning only
-- Total code reduction: ~838 lines
+- Total code reduction: ~643 lines
+
+**LLM Provider Support:**
+- Unified OpenAI client supports both Azure OpenAI and Ollama
+- Single implementation using OpenAI package's compatibility
+- Ollama uses OpenAI-compatible `/v1` endpoint
+- Configurable via `LLM_PROVIDER` environment variable
 
 **Module Improvements:**
 - Renamed `analysis.py` → `report_generator.py` (clearer purpose)
-- Renamed `azure_client.py` → `azure_openai_client.py` (clarity)
+- Renamed `azure_client.py` → `openai_client.py` (unified client)
 - Moved plan command from diary to separate `plan_commands.py` module
 - Updated `EntryManager` to accept separate diary_path and planner_path
 

README.md

@@ -139,11 +139,24 @@ Create `~/.config/brain/.env` with these settings:
 - `DIARY_PATH` - Path to Obsidian vault or markdown directory for reflection entries
 - `PLANNER_PATH` - Path to directory for daily plan files (separate from diary)
 
-**Azure OpenAI** (required for all LLM features):
+**LLM Provider** (choose one):
+- `LLM_PROVIDER` - Set to `azure` or `ollama` (default: `azure`)
+
+**Option 1: Azure OpenAI** (cloud-based, requires API key):
 - `AZURE_OPENAI_API_KEY` - Your Azure OpenAI API key
 - `AZURE_OPENAI_ENDPOINT` - Your Azure OpenAI endpoint
-- `AZURE_OPENAI_DEPLOYMENT` - Deployment name (default: gpt-4o)
-- `AZURE_OPENAI_API_VERSION` - API version (default: 2024-02-15-preview)
+- `AZURE_OPENAI_DEPLOYMENT` - Deployment name (default: `gpt-4o`)
+- `AZURE_OPENAI_API_VERSION` - API version (default: `2024-02-15-preview`)
+
+**Option 2: Ollama** (local, free, private):
+- `OLLAMA_BASE_URL` - Ollama API URL (default: `http://localhost:11434`)
+- `OLLAMA_MODEL` - Model name (default: `llama3.1`)
+- First install Ollama from https://ollama.com
+- Then pull a model: `ollama pull llama3.1`
+
+> **Switching Providers:** Just change `LLM_PROVIDER` in your `.env` file and restart. No code changes needed!
+>
+> **Privacy:** Ollama runs completely locally. Your journal entries never leave your machine.
 
 **Logging & Cost Tracking** (optional configuration):
 - `BRAIN_COST_DB_PATH` - Path to cost tracking database (default: ~/.brain/costs.db)
@@ -207,7 +220,10 @@ After writing, run `brain diary link today` to add semantic backlinks:
 
 - Python 3.13+
 - uv package manager (ALWAYS use `uv`, never `pip`)
-- **Azure OpenAI** - Required for all LLM operations:
+- **LLM Provider** (choose one):
+  - **Azure OpenAI** - Cloud-based, requires API key and subscription
+  - **Ollama** - Local, free, completely private (requires installation from https://ollama.com)
+- All LLM features supported with either provider:
   - Diary prompt generation
   - Task extraction from diary entries
   - Semantic backlinks and tags

SETUP_CHECKLIST.md

@@ -71,11 +71,18 @@ Edit your `.env` file with the following **required** variables:
 DIARY_PATH=/Users/yourname/Documents/second-brain/diary
 PLANNER_PATH=/Users/yourname/Documents/second-brain/planner
 
-# Required: Azure OpenAI (get from Azure Portal)
+# Required: LLM Provider (choose one)
+LLM_PROVIDER=azure  # or "ollama" for local
+
+# Option 1: Azure OpenAI (cloud-based)
 AZURE_OPENAI_API_KEY=your-api-key-here
 AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
 AZURE_OPENAI_DEPLOYMENT=gpt-4o
 AZURE_OPENAI_API_VERSION=2024-02-15-preview
+
+# Option 2: Ollama (local, free)
+# OLLAMA_BASE_URL=http://localhost:11434
+# OLLAMA_MODEL=llama3.1
 ```
 
 **Optional variables** (with defaults):
@@ -89,9 +96,11 @@ BRAIN_LOG_LEVEL=INFO
 BRAIN_LOG_FILE=~/.brain/logs/brain.log
 ```
 
-### Step 4: Get Azure Credentials
+### Step 4: Setup LLM Provider
+
+Choose **one** of the following:
 
-#### Azure OpenAI (Required)
+#### Option 1: Azure OpenAI (Cloud)
 
 1. Go to [Azure Portal](https://portal.azure.com)
 2. Navigate to your Azure OpenAI resource
@@ -101,6 +110,25 @@ BRAIN_LOG_FILE=~/.brain/logs/brain.log
    - **Endpoint** → `AZURE_OPENAI_ENDPOINT`
 5. Go to **Model deployments**
 6. Copy your deployment name (e.g., "gpt-4o") → `AZURE_OPENAI_DEPLOYMENT`
+7. Set `LLM_PROVIDER=azure` in `.env`
+
+#### Option 2: Ollama (Local)
+
+1. Install Ollama from https://ollama.com
+2. Start Ollama (it runs in the background)
+3. Pull a model:
+   ```bash
+   ollama pull llama3.1        # Fast, balanced (4.7GB)
+   # OR
+   ollama pull llama3.2        # Smaller, faster (2GB)
+   # OR
+   ollama pull qwen2.5:7b      # Great for structured output
+   ```
+4. Set in `.env`:
+   ```bash
+   LLM_PROVIDER=ollama
+   OLLAMA_MODEL=llama3.1  # or whichever model you pulled
+   ```
 
 ### Step 5: Test Configuration
 
@@ -136,12 +164,19 @@ The directory doesn't exist. Create it:
 mkdir -p /path/to/dir
 ```
 
-### "Azure OpenAI credentials required"
+### "Azure OpenAI credentials required" or "LLM_PROVIDER must be set"
+
+**If using Azure OpenAI:**
+1. Set `LLM_PROVIDER=azure` in `.env`
+2. Ensure `AZURE_OPENAI_API_KEY` is set correctly
+3. Ensure `AZURE_OPENAI_ENDPOINT` ends with a `/`
+4. Ensure `AZURE_OPENAI_DEPLOYMENT` matches your deployment name in Azure Portal
 
-Your Azure OpenAI credentials are missing or incorrect. Double-check:
-1. `AZURE_OPENAI_API_KEY` is set correctly
-2. `AZURE_OPENAI_ENDPOINT` ends with a `/`
-3. `AZURE_OPENAI_DEPLOYMENT` matches your deployment name in Azure Portal
+**If using Ollama:**
+1. Set `LLM_PROVIDER=ollama` in `.env`
+2. Make sure Ollama is running (check with `ollama list`)
+3. Pull a model if you haven't: `ollama pull llama3.1`
+4. Set `OLLAMA_MODEL` to match the model you pulled
 
 ### .env File Not Loading