Eva Admin

● unsaved

Select a file to edit

← Select a file from the sidebar

Changes take effect after restarting Eva.

Channels

Active message channels and their connection status

Skills

Agent skills mounted from container/skills/

API Keys

Each key is its own named channel with independent identity and permissions

Name	Key	Created	Last Used	Requests	Status	Actions

API Documentation

Base URL: http://<host>:<HTTP_API_PORT>
Auth: X-API-Key: your-key or Authorization: Bearer your-key

Text Chat — POST /api/v1/chat

curl -X POST http://host:PORT/api/v1/chat \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"message": "Turn on the living room lights"}'

Audio Chat — POST /api/v1/chat/audio

curl -X POST http://host:PORT/api/v1/chat/audio \
  -H "X-API-Key: YOUR_KEY" \
  -F "audio=@recording.mp3"

Both endpoints return {"id":"...","status":"queued"}. Responses arrive via SSE stream: GET /api/stream

Local Model

Route simple intents to a local LLM for faster, cheaper responses

How it works: Simple intents (lights, timers, time queries) are classified first. High-confidence matches go to the local model. Everything else goes to Claude. Requires llama-cpp-python or ollama running locally.

Configuration

Server Type

Server URL

Confidence Threshold (0–1)

Enable routing

Available Models

Download Model

Quick-start Instructions

Option A: llama-cpp-python

pip install llama-cpp-python[server]
python -m llama_cpp.server --model /path/to/model.gguf --host 0.0.0.0 --port 8080

Option B: Ollama

ollama pull qwen2.5:0.5b   # tiny fast model
ollama serve

GitHub Backup

Auto-commit code, memory, skills and config to a GitHub branch

GitHub Repo URL

Personal Access Token (PAT)

Branch

Enable nightly backup

Backup Time

What gets backed up: All code, memory files, skills, and config. .env secrets are redacted — keys, tokens, passwords replaced with REDACTED. Backup only runs if there are uncommitted changes.

Eva HTTP API

RESTful + SSE API for integrating with Eva programmatically. All requests require an API key.

Base URL: https://eva.lcho.me
Auth: X-API-Key: <key> or Authorization: Bearer <key>
API keys are managed in the API Keys tab.

GET /api/v1/health No auth required

Returns server status and capability flags.
Response 200: {"status":"ok","sseClients":0,"tts":true,"stt":true}

GET /api/v1/stream SSE

Server-Sent Events stream. Keep this connection open to receive Eva's responses in real time.

Events:

event: connected
data: {"clientId":"<uuid>"}

event: message
data: {"text":"Eva's reply","timestamp":"<iso8601>"}

event: audio
data: {"audio":"<base64-mp3>","mimeType":"audio/mpeg","timestamp":"<iso8601>"}

: keepalive  (every 30s)

The audio event fires alongside message when ElevenLabs TTS is configured. Decode base64 and play as MP3.

POST /api/v1/chat

Send a text message to Eva. Response arrives asynchronously via /api/v1/stream.

Body (JSON):

{"text":"What's the weather?","sender_name":"Louis"}

Response 202:

{"id":"api-1234-abcd","status":"queued"}

POST /api/v1/chat/audio

Send raw audio. ElevenLabs Scribe transcribes it, Eva replies. Response arrives via /api/v1/stream as both text and audio.

Headers:

Content-Type: audio/webm   (or audio/ogg, audio/wav, audio/mp4)
X-Sender-Name: Louis       (optional)

Body: Raw audio bytes (max 25 MB).
Response 202:

{"id":"api-1234-abcd","status":"queued","transcription":"🎤 What's the weather?"}

Quick-start (curl)

# Open stream (keep running)
curl -N -H "X-API-Key: <key>" https://eva.lcho.me/api/v1/stream

# Send text
curl -X POST https://eva.lcho.me/api/v1/chat \
  -H "X-API-Key: <key>" -H "Content-Type: application/json" \
  -d '{"text":"Hello Eva"}'

# Send voice
curl -X POST https://eva.lcho.me/api/v1/chat/audio \
  -H "X-API-Key: <key>" -H "Content-Type: audio/webm" \
  --data-binary @recording.webm

Security: All endpoints except /api/v1/health require a valid API key. Each key gets its own isolated conversation history. Manage keys in the API Keys tab.