github.com/Kocoro-lab/Shannon changelog and release notes, page 2

Shannon v0.1.0

🖥️ Desktop Apps

Download the desktop application for your platform:

macOS (Universal Binary - Intel & Apple Silicon)

.dmg installer (recommended)
.app.tar.gz for manual installation

Windows

.msi installer (recommended)
.exe NSIS installer (alternative)

Linux

.AppImage (portable, no installation required)
.deb package (Debian/Ubuntu)

🐳 Docker Images

docker pull waylandzhang/gateway:v0.1.0
docker pull waylandzhang/orchestrator:v0.1.0
docker pull waylandzhang/agent-core:v0.1.0
docker pull waylandzhang/llm-service:v0.1.0

🚀 Quick Start

Recommended: One-command install

curl -fsSL https://raw.githubusercontent.com/Kocoro-lab/Shannon/v0.1.0/scripts/install.sh | bash

This downloads all required files (config, migrations, WASM interpreter) and starts services.

Manual setup: Clone the repo and use deploy/compose/docker-compose.release.yml with the full directory structure.

📚 Documentation

📝 What's Changed

See CHANGELOG.md for detailed changes.

What's Changed

Add Claude Code GitHub Workflow by @Kocoro-lab in #3
Update web search configuration and provider support by @waylandzhang in #2
Fix: Triple bugs - duplicate session updates, timestamp accuracy, CLI error parsing by @waylandzhang in #5
Update project configuration and documentation by @KevinKanggogogo in #6
feat: Add centralized pricing, Python WASI execution, and P2P coordination by @waylandzhang in #7
feat: migrate OpenAI Chat Completions API to Responses API by @Kocoro-lab in #8
fix: Improve tool selection and parameter handling by @waylandzhang in #9
docs: Refactor deployment documentation for improved clarity and maintainability by @KevinKanggogogo in #10
feat: Add unified HTTP gateway and enhanced streaming capabilities by @waylandzhang in #11
fix: Re-enable Python WASI executor with protobuf v5 compatibility by @waylandzhang in #12
fix: Address critical workflow inconsistencies and edge cases by @waylandzhang in #14
feat: Add comprehensive observability system with event tracking and dashboard by @waylandzhang in #15
Implement Comprehensive Monitoring Infrastructure with Grafana and Prometheus by @KevinKanggogogo in #17
feat(memory): Enhanced supervisor memory system by @waylandzhang in #18
Fix P2P coordination hang and consolidate configuration system by @waylandzhang in #19
fix: Context compression with sliding window and message rehydration by @waylandzhang in #20
feat: Increase resource limits for real-world workloads by @waylandzhang in #22
feat: Add OpenAPI and MCP tool integrations with retry logic by @waylandzhang in #23
refactor: Major LLM provider system overhaul by @waylandzhang in #24
feat: Add tool error reporting to unified API responses by @waylandzhang in #25
feat: Python SDK and Provider Rate Control Improvements by @waylandzhang in #26
refactor: modernize tool handling and role preset system by @waylandzhang in #36
docs: add Chinese documentation by @AlexZander-666 in #38
feat: Add OpenAPI tool infrastructure and vendor adapter pattern by @waylandzhang in #45
fix: Improve SSE event visibility and fix OpenAPI retries closure bug by @waylandzhang in #47
feat: Improve SSE event visibility and user-friendly messaging by @waylandzhang in #48
feat: Add unified task submission endpoint with stream URL by @waylandzhang in #49
feat: Add session API endpoints and event types documentation by @waylandzhang in #50
feat: Add list sessions endpoint with pagination and type fixes by @waylandzhang in #52
fix: Add CORS support for streaming endpoints and X-User-Id header by @waylandzhang in #53
docs: Align all documentation with codebase implementation by @waylandzhang in #55
feat: Add gateway API endpoints and P0 security fixes by @waylandzhang in #56
feat: Add session soft delete endpoint by @waylandzhang in #57
feat: Add session title API with auto-generation by @waylandzhang in #58
Fix: Support dual-ID pattern for session handling by @waylandzhang in #59
feat: Add grouped-by-turn session events endpoint by @waylandzhang in #60
feat: Python SDK v0.1.0a3 - HTTP-only with idempotency and tracing by @waylandzhang in #62
fix: Comprehensive task submission parameter handling by @waylandzhang in #63
fix: Address code review findings for model migration by @waylandzhang in #64
fix: Correct provider field for all LLM providers across all workflows by @waylandzhang in #65
fix: UTF-8 string handling and environment variable configuration by @waylandzhang in #66
Improve Responses API handling for GPT-5 models by @Awakehsh in #67
examples(simple_streaming): wait for final result with timeout to avoid None by @Awakehsh in #68
feat(orchestrator): Deep Research Workflow v2 with Citation Infrastructure by @waylandzhang in #69
fix(orchestrator): Auto-persistence, per-agent costs, and comprehensive research improvements by @waylandzhang in #71
fix: UTF-8 character handling in streaming events by @waylandzhang in #73
orchestrator(roles): skip auto tool selection for role agents by @waylandzhang in #74
fix(dashboard): repair SSE streams and radar by @waylandzhang in #72
fix: resolve 2048 token limit in Deep Research sub-tasks by @Awakehsh in #75
feat: upgrade GPT-5.1 and Grok models to latest versions by @Awakehsh in #76
SSE format review - no breaking changes identified by @Copilot in #78
SSE OpenAI compatible fix by @waylandzhang in #77
feat(orchestrator): improve synthesis quality and security filtering by @waylandzhang in #80
feat(web_fetch): add subpages crawling support by @waylandzhang in #82
fix(token-tracking): complete synthesis recording and deprecate agent_usages by @waylandzhang in #84
fix(llm-service): max_tokens may exceed the allowed range allowed by llm. by @tmdhhl in #86
fix(desktop): improve ime enter handling by @Awakehsh in #87
feat(web_search): Add SerpAPI search provider support by @Awakehsh in #89
docs/windows local dev init setup guide docs by @HuangXihuang in #92
fix(makefile): clear OTEL vars during build and split build/up by @jbang2004 in #44
feat: add Google Analytics 4 (GA4) integration by @waylandzhang in #91
Feature/firecrawl fetch provider by @Awakehsh in #93
feat: add pause/resume/cancel workflow control signals by @waylandzhang in #94
Update Claude Code workflows (ANTHROPIC_API_KEY + fork PR reviews) by @waylandzhang in #97
Feat/update claude workflows by @waylandzhang in #98
fix: enforce complete tenant isolation in session operations and add refresh token rotation by @waylandzhang in #100
feat(tools): Architect Web Fetch 2.0 - Specialized Tools and Async Crawling by @Awakehsh in #101
Fix: Research Workflow Stability Improvements by @Awakehsh in #102
fix: resolve container failures in vLLM deployment setup by @waylandzhang in #103
feat: Research Infrastructure Upgrade & Deep Research V2 Robustness by @Awakehsh in #105
fix: add agent_id column to token_usage table by @waylandzhang in #106
fix: allow synthesis_model_tier override in synthesis activity by @waylandzhang in #108
feat: improve research quality - conflict resolution, citation priority, timeout handling by @Awakehsh in #109
fix: improve tool rate limiting with agent_id fallback by @Awakehsh in #110

New Contributors

@Kocoro-lab made their first contribution in #3
@waylandzhang made their first contribution in #2
@KevinKanggogogo made their first contribution in #6
@AlexZander-666 made their first contribution in #38
@Awakehsh made their first contribution in #67
@Copilot made their first contribution in #78
@tmdhhl made their first contribution in #86
@HuangXihuang made their first contribution in #92
@jbang2004 made their first contribution in #44

Full Changelog: https://github.com/Kocoro-lab/Shannon/commits/v0.1.0

Open source →

Additional notes

Added

Desktop Application

Pre-built Binaries: Native desktop apps for macOS, Windows, and Linux
- macOS: Universal binary (Intel + Apple Silicon) as .dmg and .app.tar.gz
- Windows: .msi and .exe installers
- Linux: .AppImage (portable) and .deb packages
Web UI Mode: Run as local web server (npm run dev at http://localhost:3000)
OSS Mode: Works without authentication - skip login and go directly to run page
Session Management: Conversation history with auto-generated titles
Real-time Event Timeline: Visual execution flow with SSE streaming
Multi-model Support: Switch between LLM providers in the UI
Research Agent: Deep research mode with configurable strategies

Core Features

OpenAI-compatible API: Drop-in replacement (/v1/chat/completions)
Multi-Agent Orchestration: Temporal-based workflows with DAG, Supervisor, and Strategy patterns
SSE Streaming: Real-time task execution events via Server-Sent Events
Multi-tenant Architecture: Tenant isolation, API key scoping, and rate limiting
Scheduled Tasks: Cron-based recurring task execution with Temporal schedules
Token Budget Control: Hard caps with automatic model fallback

Workflows & Research

Research Workflow: Multi-step research with parallel agent execution and synthesis
Citation System: Automatic source tracking with [n] format citations
Model Tier Override: Per-activity model tier control (e.g., synthesis_model_tier)
Synthesis Templates: Customizable output formatting via templates

Infrastructure

Agent Core (Rust): WASI sandbox, gRPC server, policy enforcement
Orchestrator (Go): Task routing, budget enforcement, OPA policies
LLM Service (Python): 15+ LLM providers (OpenAI, Anthropic, Google, DeepSeek, local models)
MCP Integration: Native Model Context Protocol support for custom tools
Python SDK: Official client library with CLI (pip install shannon-sdk)
Observability: Prometheus metrics, Grafana dashboards, OpenTelemetry tracing

Developer Experience

Hot-reload Configuration: Update config/shannon.yaml without restarts
Vendor Adapter Pattern: Clean separation for custom integrations
Release Automation: GitHub Actions builds Docker images + desktop apps on version tags

Fixed

Tool Rate Limiting: Use agent_id fallback to avoid asyncio collision
Rate Limit Accuracy: Return remaining wait time instead of full interval
API Key Normalization: Support both sk-shannon-xxx and sk_xxx formats
Citation Priority: Let CitationAgent determine source ranking dynamically
Race Conditions: Generate request UUID upfront to avoid concurrent execution issues
Research Role Support: Accept research_supervisor in decomposition endpoint

Security

WASI Sandbox: Secure Python code execution in WebAssembly sandbox
JWT Authentication: Token-based auth with refresh tokens and revocation
API Key Hashing: Store hashed API keys with prefix-based lookup
Multi-tenant Isolation: User/tenant scoping on all operations
OPA Policy Governance: Fine-grained access control rules

Documentation

Platform Guides: Ubuntu, Rocky Linux, Windows setup instructions
API Reference: OpenAPI spec and endpoint documentation
Desktop App Guide: Development, building, and troubleshooting
Vendor Adapters: Custom integration pattern documentation

Technical Details

Languages: Go 1.22+, Rust (stable), Python 3.11+
Infrastructure: Temporal, PostgreSQL, Redis, Qdrant
Desktop: Next.js, Tauri 2, React
Protocols: gRPC, HTTP/2, Server-Sent Events

github.com/Kocoro-lab/Shannon

Release timeline

Releases