What's new in
ModelPilot AI

Full analytics suite with daily bar charts, per-model cost distribution, hour-of-day message heatmap, and top conversations by token usage with sentiment scoring.

Fire webhooks on any event (conversation.ended, handoff.triggered, etc.). One-click connect to n8n, Slack, Zapier, Make, HubSpot, and Zendesk.

Generate API keys with scoped permissions (chat, analytics, logs, knowledge, webhooks). Includes built-in curl reference for all major endpoints.

Test Preview tab now calls real AI — responses stream from Claude using your configured system prompt. No more mock replies.

Added Ollama (local), Groq (free tier), OpenRouter (free models), Together.ai ($25 credits), HuggingFace, Mistral, Fireworks, and Cohere — all with LiteLLM config snippets.

Groq + OpenRouter free models + Ollama + Cohere embeddings = fully functional RAG chatbot at zero cost.

FAQ, Support, Sales, Language, Onboarding, and Human Handoff agents — each configurable with retrieval strategy, confidence threshold, and advanced toggles.

Sprint plan (6 weeks), DB schema (6 tables with RLS), code snippets (chat endpoint, RAG pipeline, widget), and full architecture overview.

KPI cards with sparklines, 14-day area chart, bot activity donut, and a live conversation feed table.

Upload PDFs, scrape URLs, edit FAQs, and configure chunking strategy — all in one screen.

Filter by bot, status, and search by user. Click any row to see the full transcript, metadata, and cost breakdown.

Full modal editor with Config, Personality, and Test Preview tabs. Temperature slider, model selector, feature toggles (handoff, web search, lead capture).

Customise accent colour, bot name, greeting, and position. See changes reflected in the live widget preview instantly. One-click copy embed code.

Welcome → Workspace → API Key → Create Bot → Upload Knowledge → Embed. Progress bar and animated transitions.