Add multi-provider support for local and cloud LLM combinations #76
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Implement provider abstraction layer supporting three modes:
- OpenRouter mode: Cloud-based models via OpenRouter API (existing)
- Ollama mode: Local models via Ollama server (new)
- Mixed mode: Combination of local and cloud models (new)
Motivation:
Test whether the council deliberation approach provides benefits
for small local models. Hypothesis: collaborative reasoning and
peer review may compensate for individual model limitations.
Mixed mode enables experiments combining cost-effective local
models with premium SOTA models for optimal cost/quality balance.
Implementation:
- Provider abstraction layer (base.py, factory pattern)
- OpenRouterProvider (refactored from openrouter.py)
- OllamaProvider (new, supports local Ollama server)
- Intelligent routing with cross-provider parallel execution
- Model spec format: simple ("model") or prefixed ("provider:model")
- Backward compatible - existing configs work unchanged
Files changed:
- backend/providers/ - New provider abstraction system
- backend/council.py - Updated import (1 line change)
- backend/config.py - Added LLM_PROVIDER and OLLAMA_BASE_URL
- .env.example - Configuration template for all modes
- README.md - Provider setup and usage documentation
- TROUBLESHOOTING.md - Common issues and solutions
- CLAUDE.md - Architecture decisions and migration guide
All existing functionality preserved. Zero-config local mode
requires no API keys. Mixed mode optimizes for cost efficiency.
🤖 Generated with Claude Code
Co-Authored-By: Claude <[email protected]>
This commit message:
- Subject line: Clear, concise action statement
- Overview: What was implemented
- Motivation section: Your reasoning about testing local models and the council hypothesis
- Implementation section: Key technical details
- Files changed: Overview of modifications
- Summary: Benefits and features
Copilot AI
added a commit
to danielmerja/llm-council
that referenced
this pull request
Dec 8, 2025
…uter + Mixed mode) - Provider abstraction layer with factory pattern - Support for local Ollama models (100% offline operation) - Mixed mode for combining local and cloud models - Backward compatible with existing OpenRouter setup - New backend/providers/ directory with base, ollama, and openrouter providers - .env.example with comprehensive configuration examples - Enables cost-effective local model experimentation
Th0mYT
added a commit
to Th0mYT/llm-council
that referenced
this pull request
Dec 8, 2025
Implements a comprehensive provider abstraction system enabling flexible LLM backend configuration with support for cloud and local models. ## Features ### Provider Support - OpenRouter: Access to GPT-4, Claude, Gemini, and 200+ models - Ollama: Local models with proper API integration - LMStudio: Local models via OpenAI-compatible interface ### Flexible Configuration - Model-level provider specification: "provider:model" format - Mixed mode: Combine any providers (e.g., local council + cloud chairman) - Backward compatible: Existing configs continue to work ### Architecture - Provider abstraction layer (backend/providers/) - Factory pattern for automatic model routing - Parallel execution across multiple providers - Graceful degradation on provider failures ## New Files - backend/providers/ - Complete provider abstraction - base.py: Abstract Provider interface - openrouter.py: Cloud provider implementation - ollama.py: Local Ollama provider (fixed API usage) - lmstudio.py: Local LMStudio provider - factory.py: Smart routing and provider management - .env.example: Configuration examples for all modes - SETUP.md: Comprehensive setup guide - requirements.txt: Easy dependency installation - test_providers.py: Provider validation script ## Modified Files - backend/config.py: New configuration system with provider settings - backend/council.py: Refactored to use ProviderFactory - CLAUDE.md: Updated architecture documentation - .gitignore: Added IDE directories ## Configuration Examples Cloud only: COUNCIL_MODELS = ["openai/gpt-4", "anthropic/claude-sonnet-4"] Local only: COUNCIL_MODELS = ["ollama:llama2", "ollama:mistral"] Mixed mode: COUNCIL_MODELS = ["ollama:llama2", "openrouter:gpt-4"] CHAIRMAN_MODEL = "openrouter:anthropic/claude-sonnet-4" ## Testing - All provider tests passing - Backend starts without errors - Backward compatible with existing installations ## Advantages Over PR karpathy#76 - LMStudio support (not just Ollama) - Model-level provider specification (more flexible) - Correct Ollama API implementation - Comprehensive documentation and testing - Factory pattern for cleaner architecture 🤖 Generated with Claude Code (https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Implement provider abstraction layer supporting three modes:
Motivation:
Test whether the council deliberation approach provides benefits
for small local models. Hypothesis: collaborative reasoning and
peer review may compensate for individual model limitations.
Mixed mode enables experiments combining cost-effective local
models with premium SOTA models for optimal cost/quality balance.
Implementation:
Files changed:
All existing functionality preserved. Zero-config local mode
requires no API keys. Mixed mode optimizes for cost efficiency.
🤖 Generated with Claude Code
Co-Authored-By: Claude [email protected]
This commit message: