- Fixed module path from github.com/deepblackcloud/bzzz to github.com/anthonyrawlins/bzzz
- Added dynamic Ollama model detection via /api/tags endpoint
- Implemented intelligent model selection through N8N webhook integration
- Added BZZZ_MODEL_SELECTION_WEBHOOK environment variable support
- Fixed GitHub assignee issue by using valid username instead of peer ID
- Added comprehensive model fallback mechanisms
- Updated all import statements across the codebase
- Removed duplicate systemd service file
- Added sandbox execution environment and type definitions
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Replace stub functions with real capability-based filtering
- Add task matching based on labels and keywords
- Implement P2P task claim announcements
- Support code-generation, debugging, testing, documentation tasks
- Add intelligent task type detection from issue content
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
This commit introduces the human escalation feature, completing Phase 3 of the development plan.
The agent can now detect when a conversation is unproductive and call for human help.
Key changes:
- A trigger condition has been added to detect keywords like 'stuck' or 'help', or if a conversation exceeds 10 messages.
- When triggered, the agent sends the full task context and conversation history to the N8N webhook for human review.
- An 'IsEscalated' flag is set on the conversation to prevent further automated responses from any agent in the mesh.
- A final message is broadcast to the Antennae channel to inform all peers of the escalation.
- Add Conversation struct to track task discussion history
- Implement handleMetaDiscussion for dynamic peer collaboration
- Enhanced GitHub integration with active discussion management
- Add SetAntennaeMessageHandler for pluggable message handling
- Simplify pubsub message types to generic MetaDiscussion
- Enable real-time collaborative reasoning between AI agents
- Integrate conversation context into Ollama response generation
- Support distributed decision making across P2P network
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
This commit introduces the new 'reasoning' module, which allows the bzzz-agent to connect to a local Ollama instance.
It also modifies the 'github' integration module to use this new capability. When a task is claimed, the agent now generates a step-by-step execution plan using an Ollama model and publishes this plan to the Antennae meta-discussion channel for peer review before proceeding.
This completes Phase 1 of the Ollama integration plan.
