AI Support Agent - Ragrails

The problem

Support teams answer the same questions repeatedly. Generic chatbots make the problem worse when they invent refund rules, pricing terms, or policy exceptions that are not in your docs.

Why Ragrails fits

Ragrails gives you a grounded support workflow: ingest help content, retrieve the right chunks, rewrite follow-ups, and tune low-confidence behavior so the bot can answer carefully or refuse.

Need	Ragrails capability
Help center plus policy files in one index	Extraction and Pipeline Overview
Multi-turn support questions	Query Rewriting
Avoid hallucinated policy answers	Chat Tuning
Stream a responsive support UI	Streaming
Keep policies current	Knowledge Base Maintenance

Sample

from ragrails import ChatRetrievalQualityConfig, QueryRewriteConfig, RagRails

rag = RagRails(
    collection="support",
    vector_store={"provider": "qdrant", "url": "http://localhost:6333"},
    embedding={"provider": "voyage", "model": "voyage-3"},
    llm={"provider": "openai", "model": "gpt-4o-mini"},
)

rag.ingest(
    urls="https://help.example.com",
    docs={"folder": "files/policies/"},
    ingestion={"urls": {"mode": "full", "max_pages": 200}},
    storage={"collection": "support"},
)

result = rag.chat(
    "Can I get a refund after 30 days?",
    history=[],
    query_rewrite=QueryRewriteConfig(enabled=True),
    retrieval_quality=ChatRetrievalQualityConfig(
        low_confidence_mode="refuse_grounded_answer",
        min_retrieval_score=0.35,
    ),
)

print(result.answer)
print(result.answer_confidence)

ragrails setup-url --browser chromium

ragrails ingest \
  --source-url https://help.example.com \
  --folder files/policies/ \
  --vector-db qdrant \
  --collection support \
  --url http://localhost:6333 \
  --provider voyage \
  --model voyage-3

ragrails chat "Can I get a refund after 30 days?" \
  --rewrite-query \
  --vector-db qdrant \
  --collection support \
  --url http://localhost:6333 \
  --embedder-provider voyage \
  --embedder-model voyage-3 \
  --llm-provider openai \
  --llm-model gpt-4o-mini

curl -X POST http://127.0.0.1:8000/v1/pipelines/ingest \
  -H "Content-Type: application/json" \
  -d '{
    "urls": "https://help.example.com",
    "docs": {"folder": "files/policies/"},
    "ingestion": {"urls": {"mode": "full", "max_pages": 200}},
    "embedding": {"provider": "voyage", "model": "voyage-3"},
    "storage": {"vector_db": "qdrant", "collection": "support", "url": "http://localhost:6333"}
  }'

curl -X POST http://127.0.0.1:8000/v1/chat \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Can I get a refund after 30 days?",
    "query_rewrite": {"enabled": true},
    "retrieval_quality": {"low_confidence_mode": "refuse_grounded_answer", "min_retrieval_score": 0.35},
    "llm_provider": "openai",
    "llm_model": "gpt-4o-mini",
    "embedder_provider": "voyage",
    "embedder_model": "voyage-3",
    "vector_db": "qdrant",
    "collection": "support",
    "url": "http://localhost:6333",
    "history": []
  }'

Production notes

Persist result.history per customer session.
Display result.sources so agents and customers can inspect the policy source.
Use answer_confidence and retrieval_quality to route weak answers to a human.
Refresh changed policies with Knowledge Base Maintenance.