Flash documentation
Orientation for Teralor Flash: what it is, how to get access, how to connect your tools, and how the API behaves.
Flash is Teralor's fast model for rapid task completion. Access is authorized-customer only: one
HTTPS base endpoint, one product name (Flash), and one bearer key issued through your account.
These pages describe the current public surface: the endpoints, fields, and statuses the API
actually serves today.
How the pieces fit
- You (the primary model or human) decide scope, architecture, and final acceptance.
- Flash executes bounded tasks fast: generation, transformation, analysis, and tool-call proposals.
- Your harness (Hermes, Pi, Cursor, ZCode, Claude Code, Codex, or a plain HTTP client) is the only thing that executes local tools and applies changes. Flash proposes; your harness validates and executes.
Where to go next
- Start here — approved access, key creation, environment setup, and your first completion.
- Connect tools — supported roles per harness (primary and/or executor), with the canonical downloadable setup skills.
- Use Flash well — primary vs executor roles, supervised scope, streaming, tool execution, and context budgeting.
- API reference — the supported routes (
GET /v1/models,POST /v1/chat/completions), fields, streaming, and errors. - Limits and billing — what decides access today, and where to see your own plan, allowance, and usage.
- Troubleshooting — status codes, context failures, interrupted streams, and how to file a useful support request.
Authority
The live API is the authority. Authenticated GET /v1/models returns the model catalog with
the current limits and capabilities, and these pages defer to it wherever they overlap. Marketing
context lives on the public pages (/api, /executor, /orchestrator);
this docs tree is the operational reference.