OpenChat-1openchat.one

Technical documentation

OpenChat-1

LLM + real data = real insights

OpenChat is not another chat wrapper. It is a minimal interface over multi-model LLMs (Grok by default via OpenRouter/xAI, or any model you choose) permanently wired to a multi-sector historical and realtime data lake — the leverage that turns fluency into facts.

Thesis

Language models alone invent plausible numbers. Data platforms alone dump tables without narrative. OpenChat multiplies both: the model reasons; the lake constrains. Every conversation can call catalog search, dataset metadata, historical JSONL packages, and realtime feeds. Citations use stable dataset slugs.

insights = llm.reason(question, tools=openchat_data)
// tools: search_catalog, get_dataset_records,
//        list_realtime_feeds, get_platform_overview, …

Architecture (open infrastructure)

  • Chat surface — minimal UI at /chat
  • Model router — OpenRouter multi-model + direct xAI Grok fallback
  • Data tools — first-party connector to the OpenChat lake (biggest leverage)
  • Agent templates — product research, IG+Replicate+Blotato, Anthropic docs, trading bots, …
  • Connectors — API-to-API open definitions anyone can publish
  • Credits — prepurchase tokens; 1 free prompt after signup + captcha

Source for built-ins lives under apps/web/src/lib/openchat/ — templates, connectors, catalog. Community plugins are stored and listed in the library; the goal is continuous growth without a closed marketplace silo.

Agent templates (examples)

  • Demo Factory (first use-case) — Playwright records OpenChat + ElevenLabs voice + Blotato 3 videos/day on X & Instagram. See /docs/demo-pipeline
  • Product research — competitors, pricing, demand signals grounded in lake data
  • Instagram video posts — script → Replicate video → Blotato publish (bring API keys)
  • Document writer — Anthropic/Claude long-form with dataset citations
  • Trading bots — upload code + equity history; co-pilot with market feeds

Browse /library, publish at /publish, trading at /trading.

Billing

Users sign up with captcha, receive one free prompt, then prepurchase token credit packs. Each chat completion settles prompt+completion tokens against the wallet. Platform LLM keys (OpenRouter/xAI) are server-side; user connector keys (Replicate, Blotato, Anthropic direct) stay with the user for specialized agents.

Growing the ecosystem

  1. Add a connector under lib/openchat/connectors/
  2. Add an agent template under lib/openchat/templates/
  3. Register in catalog.ts or publish via UI
  4. Ship more lake packages so every agent gets smarter for free

Domain: openchat.one

Open infrastructure guide →

Technical docs — LLM + real data = real insights | OpenChat-1