Skillget

RideRecall-Agent

mcp
Guvenlik Denetimi
Gecti
Health Gecti
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 19 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Stateful ride-support agent: versioned claims, evidence-backed RAG, durable memory and guarded MCP tools.

README.md

RideRecall-Agent

Verify

A stateful ride-support agent with durable memory, versioned claims and evidence-grounded replies. Python · FastAPI · Vue · SQLite · Chroma · MCP · Ollama.

01 Choose the charged trip; 02 Correct the account; 03 Inspect policy evidence

04 Hand off the whole case; 05 Confirm the adjustment; 06 Close the loop

Click either row to enlarge it. All 25 screens.

Architecture

Request processing, policy and trip lookups, case storage and adjustment confirmation

  • Durable memory: restores conversations and claim revisions while keeping rider statements separate from trip facts.
  • Evidence-grounded analysis: policy RAG and read-only MCP supply evidence; local Qwen3 produces structured corrections and response plans.
  • Guarded case actions: specialist handoff, explicit rider confirmation, revision checks and atomic SQLite writes with a Chroma outbox.

Evaluation

Local Qwen3 4B · full API, memory, Chroma and MCP pipeline · 100 synthetic dialogues:

Metric Result
Behavioral contract checks (behavior-v2) 100/100 dialogues
Context retention 644/644 checkpoints
Reply-source audit 682/682 analyses
Analysis latency, P50 / P95 11.19 / 13.82 s

In the frozen deterministic comparison across 2,000 dialogues, durable memory achieved 100% workflow completion vs. 67% with four recent rider rounds. Benchmark · Protocols and results.

Run locally

Requires Python 3.11/3.12 and Node.js compatible with Vite 7. The default demo runs without an API key; local inference uses Ollama.

git clone https://github.com/lucia-huang7/RideRecall-Agent.git
cd RideRecall-Agent
bash scripts/dev/one-click.sh start --no-open

Open Trips, or double-click Start RideRecall-Agent.command on macOS. Setup, tests and model configuration.

Java backend reference: Java 21, Spring Boot and LangChain4j.

Trips, policies and adjustments are simulated; evaluation scope and deployment requirements are documented in Operations.

License

MIT.

Yorumlar (0)

Sonuc bulunamadi