Chatbot integrations
Put a private LLM chatbot in your product
Each ForgeStack stack is a RAG chatbot backend running in your own AWS account. Your app calls its chat API, answers come back grounded in your documents with citations, and when the model can’t help, the conversation is handed to a person in the tool your team already uses.
1. Call the chat API from your app
A live stack exposes POST /api/{projectId}/chat, authenticated with an x-api-key header. Keys are scoped per endpoint (chat, documents, train), so the key in your website backend can chat and nothing else. Responses are chat.completion-shaped, so existing parsing code carries over, and each answer lists the documents it drew on.
curl -X POST https://forgestack.au/api/<projectId>/chat \
-H "x-api-key: fsk_..." -H "content-type: application/json" \
-d '{"messages":[{"role":"user","content":"What is our refund window?"}]}'Full reference, errors and retries: docs.
2. Feed it your knowledge
Documents are chunked and indexed inside the stack, in its own S3 bucket. Sources:
- File upload or pasted text
- GitHub
- Confluence
- Jira
3. Hand off to a human
When the model can’t answer from your knowledge base, or the visitor asks for a person, the bot offers “Connect to a human?”. Nothing is sent unless the visitor accepts. After handoff the bot stops answering in that conversation, so your agent isn’t talked over. Pick one destination per project:
| Destination | What you need |
|---|---|
| Slack | a Slack incoming webhook URL |
| Microsoft Teams | a Teams incoming webhook URL |
| Zendesk | your subdomain, an agent email and an API token |
| Freshdesk | your subdomain, a requester email and an API key |
| Intercom | an Intercom access token |
| an HTTPS relay URL and the recipient address | |
| Generic webhook (JSON) | an HTTPS URL, with an optional signing secret |
The chat response reports handoff_offered and handed_off, so your UI can show the state. Delivery runs through the stack’s router on AWS.
What runs where
The model, documents, vector index and conversation logs are stored in your AWS account. Chat requests pass through ForgeStack, which checks the API key and forwards them to your stack. ForgeStack does not provide an embeddable website widget today: you call the API from your own frontend or backend.
Pick a model to start from in the model catalog, or see how this compares with Amazon Bedrock.
Deploy it into your own AWS account
Free to prototype. You pay AWS directly, and the CloudFormation template is yours to keep.