v.5.1
A lightweight web UI for chatting with multiple AI providers. No frameworks, no build steps — just HTML, CSS, and JavaScript.
Best experience: Select Eva (AIG) from the model dropdown — persistent memory, emotion tracking, proactive data retrieval, and intelligent cross-model orchestration.
AIG (Artificial Intelligence Gateway) is Eva's multi-agentic orchestration layer. Rather than calling a single model directly, AIG acts as an intelligent router and coordinator:
- Model routing — Selects the best provider for each request: GitHub Models API (GPT-4.1, GPT-5, o4-mini, DeepSeek-R1) via PAT, or Copilot CLI via ACP (Claude, Gemini, GPT-5.x) as fallback.
- MCP tool use — The underlying Copilot agent autonomously invokes Model Context Protocol servers to retrieve live data before generating a response:
- Kusto MCP — Query Azure Data Explorer for Eva's memory tables (conversations, reflections, emotion state, knowledge graph).
- GitHub MCP — Search repos, issues, and PRs via the GitHub MCP server (Docker).
- Azure MCP — Access Azure services and resources.
- Cognition pipeline — After each exchange, AIG automatically extracts knowledge entities, updates emotion vectors, logs conversations to Kusto, and triggers periodic self-reflections.
- Memory injection — Before every response, AIG fetches relevant context from Kusto (recent conversations, reflections, emotion baseline, knowledge) and injects it into the system prompt.
- Hot-reloadable MCP — Add or remove MCP servers at runtime via
POST /v1/mcp/configurewithout restarting the bridge.
The bridge (tools/acp_bridge.py) exposes an OpenAI-compatible API so any client that speaks Chat Completions can use the full AIG stack.
- Copy
config.example.json→config.jsonand add your API keys. - Open
index.htmlin a browser. - (Optional) For the full Eva experience with persistent memory:
Then select Eva (AIG) from the model dropdown.
# Seed the Kusto database (see tools/eva_seed.kql) python3 tools/acp_bridge.py --port 8888 \ --enable-kusto-mcp --kusto-cluster "https://your-cluster.region.kusto.windows.net" \ --kusto-database Eva
Tip: For local file:// usage, copy
config.local.example.js→config.local.jsinstead.
| Provider | Models | Auth |
|---|---|---|
| Eva (AIG) ⭐ | Intelligent orchestration (GPT-4.1 + ACP fallback) | Copilot CLI + optional GitHub PAT |
| OpenAI | GPT-4o, o1, o3-mini, GPT-5-mini | OpenAI API Key |
| GitHub Copilot (PAT) | GPT-4o, GPT-4o Mini, o3-mini, GPT-4.1, GPT-5, o4-mini | GitHub PAT |
| GitHub Copilot (ACP) | Claude, GPT-5.x, Gemini 3 Pro | Copilot CLI |
| Google Gemini | Gemini 2.0 Flash Thinking | Gemini API Key |
| lm-studio | Any local model | None |
| DALL-E 3 | Image generation | OpenAI API Key |
- Eva (AIG) — multi-agentic gateway with persistent memory, emotion tracking, and proactive data retrieval
- MCP tool ecosystem — Kusto (memory/analytics), GitHub (repos/issues), Azure (cloud resources) — hot-reloadable at runtime
- Cognition layer — automatic knowledge extraction, morning reflections, emotion vectors, conversation logging
- Cross-model orchestration — routes OpenAI models via GitHub Models API, Claude/Gemini via Copilot CLI ACP
- Session explorer — auto-save/restore conversations across page refresh
- Voice activation — wake-word "Eva" with continuous listening
- Settings panel with tabbed UI (General, Models, Auth, Prompts, MCP)
- Inline images — Wikimedia search or DALL-E generation
- LCARS theme — Star Trek-inspired interface
- Text-to-speech — Amazon Polly and Bark TTS
- Conversation memory in localStorage
- Database seed —
tools/eva_seed.kqlfor quick public setup - CI pipeline — secret scanning, syntax checks, model routing validation
python3 tools/acp_bridge.py --port 8888
python3 tools/acp_bridge.py --port 8888 --enable-kusto-mcp --kusto-cluster "https://..."Requires: copilot CLI installed + authenticated. Details →
See README-2.md for architecture, dependencies, and technical details.
See .github/copilot-instructions.md for contribution guidance.
Based on CodeProject. Complete overhaul.