AI Platform — Employee Portal

API endpoint

Preset payload (presets.json)

LLM provider (per call)

Request body (JSON)

Response

—

LLM provider

Temperature

JSON

Prompt

Response

—

LLM provider

System preset

System prompt (editable)

presets.json — edit and Save to persist on the server

Saving writes back to the server's presets.json (persists across restarts). The Home tab reloads its endpoints after a successful save.

LLM provider (agent brain)

System preset

Hands-free (beta)

System prompt (editable)

Push-to-talk: hold the button, speak, release. Needs mic permission + HTTPS (the mic is blocked on plain http://IP). STT = Groq Whisper · TTS = Groq Orpheus.

Mailbox (IMAP)

Count

LLM provider (agent brain)

Reads via IMAP from a configured mailbox (add in /admin); classification uses your LLM gateway.

From (SMTP account)

To

Subject

AI draft (optional) — describe the email

LLM provider (agent brain)

Body

Sends via a configured SMTP account (add in /admin).

Image (png / jpg)

LLM provider (vision brain)

Mode

Instruction (optional)

Target fields / shape (optional, for Extract)

Result

Uses a vision-capable model (default gemini; gpt-4o also works). Photograph a rate confirmation / BOL / invoice → structured JSON. Metered in the Usage tab.

Collection

New collection

Ingest

Upload file (pdf / txt / md / csv)

…or paste text

Ask

Question

LLM provider (answerer)

Answer

Embeddings run locally (ollama nomic-embed-text, free); stored in Postgres + pgvector. The answer is grounded in your documents with [n] citations, and metered in the Usage tab.

🤖 Employee Portal

Response

Response

Ingest

Ask

Flows

Agents