HezaoHezao
Featured
poirot
Poirot is a deep research agent kernel with a middleware-first architecture, five-layer long-term memory, multi-agent orchestration, sandbox isolation, and a three-layer skill self-evolution system, backed by 2400+ tests.
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It provides rule-based root cause analysis with confidence scores and redacted diagnostic export, all without mutating runtime state, and supports both replay (no backend) and connected modes.
Who it is for
- Developers building multi-agent systems with DeerFlow
- Engineers debugging complex agent failures and dependencies
- DevOps or SRE teams needing observability into agent runtimes
- Researchers or educators demonstrating agent diagnostics
Use cases
- Identifying the primary cause and contributing factors of a multi-agent run failure
- Inspecting a normalized execution timeline across all agent phases
- Replaying seven predefined failure scenarios without any backend or API keys
- Exporting redacted diagnostic data for offline analysis or LLM explanation
Strengths
- Deterministic failure attribution using rule-based evidence, with confidence and suggested actions
- Read-only diagnostics path that never modifies checkpoints, prompts, or runtime state
- Comprehensive privacy and security measures including redaction of secrets and bounded tool content
- Verified performance benchmark: median 24.8 ms for 2,000 events (local Python processing)
Considerations
- Focused on developer diagnostics in v0.1; not a production alert manager or billing platform
- Connected mode requires a fully configured DeerFlow runtime with authentication and model providers
- Replay data is synthetic and may not reflect real-world failure patterns
README quick start
Quick Start
Description
Observable diagnostics, failure attribution, metrics, and no-key replay for multi-agent runtimes.
Related repositories
Similar projects matched by category, topics, and programming language.
S40911120
Featured
recensa
Recensa is a self-hosted web viewer that indexes Claude Code session transcripts into a local SQLite database, enabling full-text search, replay, and audit of all past agent conversations without uploading data anywhere.
lopopolo
Featured
harness-engineering
Harness Engineering is a methodology for improving coding agent outputs by carefully crafting the environment around them—providing curated context, tools, and executable constraints that encode an organization’s nonfunctional requirements and cumulative lessons.