FAQs

General

Q: What is the Fastino Personalization API?
The Personalization API provides user-level memory and behavioral modeling — allowing agents to register users, ingest signals, retrieve summaries, and ground conversations in personalized context.

Q: Is the API idempotent?
Yes. Calls such as PUT /register and POST /ingest are idempotent where practical. Re-sending the same event or user data will merge or update existing records.

Q: Do I need to retrain my model after integrating?
No. Fastino runs as an independent personalization layer that works with your existing LLMs or tools. It provides structured user context that can be inserted directly into model prompts or reasoning inputs.

Q: What kind of data can I send to Fastino?
You can send structured events and documents (emails, notes, logs, or messages). The API supports both text-based and metadata-rich payloads in JSON format.

Authentication & Access

Q: How do I authenticate requests?
All API requests require a Bearer token in the header:

Authorization: x-api-key: <token>

Q: Can I have multiple API keys per workspace?
Yes. Each workspace can issue multiple scoped keys with different permissions (read-only, write-only, or full access).

Q: What permissions do keys control?
Each key is associated with a scope:

• read — allows /query and /summary

• write — allows /register and /ingest

• admin — includes deletion or configuration endpoints

Users & Data

Q: How do I register a new user?
Use:

Provide a user_id and optional traits such as name, timezone, or notes.

Q: How do I delete all data for a user?
Send:

This permanently removes all data linked to that user.

Q: How does deduplication work?
When ingesting events, set "options": {"dedupe": true}.
Fastino uses event and document IDs to automatically skip duplicates.

Q: How fresh is summary data?
Summaries are updated automatically whenever new signals arrive and can be refreshed on-demand with:

GET /summary?user_id=<id>&purpose=<purpose>

Performance & Limits

Q: What are the API rate limits?
Default limits:

• Live: 600 requests/minute
  Headers such as X-RateLimit-Remaining and Retry-After are provided for rate management.

Q: How large can a single ingestion payload be?
Up to 32 KB of content per record (event or document).
We recommend batching up to 100 events or 25 documents per request for best performance.

Q: How fast is retrieval?
Typical response times:

• /summary: ~150–300 ms (cached)

• /chunks: ~400–800 ms depending on embedding depth

• /query: ~3000 ms

Privacy & Compliance

Q: Is Fastino GDPR compliant?
Yes. All user data is encrypted at rest (AES-256) and in transit (TLS 1.3).
You can delete data globally via /delete.

Q: How long is user data retained?

• indefinite (until deleted by user or admin)

Q: Can users export their data?
Yes. Use the dashboard or contact support to export JSON-formatted user data for compliance or debugging.

Debugging & Errors

Q: How are errors returned?
All errors follow a structured schema:

Q: What should I do if I hit rate limits?
Respect the Retry-After header and back off for the specified duration.
Consider caching frequent queries or summaries.

Q: What does “USER_NOT_FOUND” mean?
The specified user_id doesn’t exist in your workspace.
Re-register the user with PUT /register.

Integrations & Agents

Q: Can Fastino integrate with LLM frameworks?
Yes. Fastino connects easily with frameworks such as LangChain, LlamaIndex, MCP, and Realtime APIs.
Use the /chunks endpoint to retrieve personalized memory for model grounding.

Q: How can I use Fastino inside ChatGPT or Claude?
Fastino exposes MCP-compatible tools (search_or_ask, update_memory, retrieve_relevant_chunks) that connect directly to ChatGPT and Claude’s MCP runtime.

Q: Can multiple agents share the same user memory?
Yes. All agents under a workspace share a unified world model per user.
This enables cross-tool reasoning, scheduling, and context handoff.

Security

Q: How does Fastino handle API key security?
Keys are scoped, revocable, and never logged in plaintext.
Use environment variables or secure vaults to manage credentials.

Q: Can I limit API key usage to certain endpoints?
Yes — you can create keys with endpoint-specific scopes (read, write, or admin).

Q: Does Fastino store raw user content?
Raw text is stored only as needed for embeddings and summarization.
All embeddings are anonymized and cannot reconstruct the original data.

Q: What encryption does Fastino use?
AES-256 at rest, TLS 1.3 for transit, and pseudonymized key-based indexing.

Support & Contact

Q: How do I get help?

• Review the API Reference for syntax and schema examples.

• Check the Quick Start for integration guides.

• Reach out to support@fastino.ai for technical or account-related issues.

Q: How do I report a bug?
Open an issue through the dashboard or email support@fastino.ai with the request_id and timestamp from the error response.

Q: Does Fastino offer SLAs or enterprise support?
Yes. Enterprise and Partner tiers include guaranteed uptime, dedicated Slack channels, and extended rate limits.