Loading…
Platform
A workspace to chat over your documents, an OpenAI-compatible API to build on them, and a dashboard that makes the reuse — and the savings — legible. Run it hosted on lab358 Cloud, or in your own AWS account.
A streaming chat UI over your indexed documents, with the model and preset picker and a cache-hit / time-to-first-token readout on every message.
Upload once, precompute the index once — then reuse it across every chat and agent. Versioned, access-controlled per document, with a readiness status (Indexed / Indexing / Queued / Failed) so you can see what's live.
Save a system prompt plus model and document configuration as a preset, then apply it across the Playground and the API so behaviour is consistent.
An OpenAI-compatible completions API, TypeScript and Python SDKs, and a command-line client — point your existing tools at lab358 and go.
Define functions and let the model call them, or constrain a response to a JSON schema so it is valid by construction rather than by hope. Both work through the same OpenAI-compatible request shape your existing code already sends.
Temperature, top-p, top-k and min-p; stop sequences; repetition, presence and frequency penalties; per-token logit bias; log-probabilities; and a seed for a repeatable draw. Requests you don't tune follow the model's own shipped defaults instead of a house preference.
On models that support it, the model can think before it answers — and the reasoning comes back on its own channel, separate from the answer, so you can show it, log it, or drop it without parsing it out of the response.
A live dashboard: tokens in and out, reuse rate, and cost, broken down by API key, model, and preset — so the efficiency win is legible.
Invite members with roles, mint and scope API keys, and (on Business) sign in with SSO. Every config change lands in an audit trail that ships continuously to your own SIEM or S3 bucket — no manual export, no gaps.
A CLOSER LOOK
Upload a document once and index it once. A readiness Status shows when it's live and ready to reuse, and every chat, agent, and API call draws on the same index — versioned and access-controlled per document.
Save a system prompt together with a model and document configuration as a preset, then apply it across the Playground and the API so behaviour stays consistent everywhere you use it.
Invite members with roles, mint and scope API keys, and sign in with SSO on Business. Every configuration change is captured in an exportable audit log.
API & SDK
lab358 speaks the OpenAI API, so the SDKs and tools you already use just work — point them at the lab358 base URL.