AskAi

The open knowledge platform for humans and AI agents. Ingest documents once. Query via chat, REST, MCP, or SDK. Deploy fully open-source, on Azure, on AWS, on-prem, or hybrid.

License: Apache 2.0 FaastLab.Ai

---

What is AskAi?

AskAi is FaastLab's flagship open-source platform for document intelligence. It does four things, all production-grade, all modular:

1. Indexing — ingest documents from any source, parse, chunk, embed
2. Search & RAG — hybrid retrieval (vector + keyword + metadata) with reranking
3. Summarisation — map-reduce summaries for documents of any length
4. Ask AI — chat over your documents, with citations and multi-step reasoning

It is designed to serve three audiences from one codebase:

• Humans — via a chat UI (Ask AI)
• AI agents — via REST, MCP server, or Python SDK (knowledge layer for agentic AI)
• Validators — agents that read documents and validate them against a corpus (e.g. regulatory report compliance)

---

Why AskAi exists

Most open-source RAG projects are demos. Most enterprise knowledge platforms are SaaS-only or vendor-locked. There's a gap in the market for an open, modular, agent-ready knowledge platform that companies can:

• Self-host fully open-source
• Deploy on Azure / AWS / GCP using managed services
• Mix and match (e.g. OpenAI for LLM, Postgres for vectors, S3 for storage)

AskAi fills that gap. Every external dependency is behind an adapter — swap OpenAI for Azure OpenAI, swap Postgres for Azure AI Search, swap MinIO for S3 — without changing application code.

---

Architecture at a glance

┌────────────────────────────────────────────────────────────────┐
│                    Interfaces (pick any/all)                    │
│   Chat UI (Next.js) │ REST API │ MCP Server │ Python SDK        │
└────────────────────────────────────────────────────────────────┘
                                │
┌────────────────────────────────────────────────────────────────┐
│                       Core Modules                              │
│  ┌──────────┐  ┌────────────┐  ┌──────────────┐  ┌──────────┐  │
│  │ Indexing │→ │ Search/RAG │→ │ Summarisation│→ │ Ask AI   │  │
│  └──────────┘  └────────────┘  └──────────────┘  └──────────┘  │
└────────────────────────────────────────────────────────────────┘
                                │
┌────────────────────────────────────────────────────────────────┐
│                    Adapter Layer (swappable)                    │
│  Storage │ Vector DB │ LLM │ Embeddings │ Queue │ Auth          │
└────────────────────────────────────────────────────────────────┘
                                │
┌────────────────────────────────────────────────────────────────┐
│       Default OSS stack (works out of the box)                  │
│  MinIO │ Postgres+pgvector │ OpenAI │ Redis+Celery │ JWT        │
│                                                                 │
│       Azure stack (one config change)                           │
│  Blob │ Azure AI Search │ Azure OpenAI │ Service Bus │ Entra    │
└────────────────────────────────────────────────────────────────┘

See Architecture.md for the full design.

---

Quick start

# Clone and enter
git clone https://github.com/faastlab-ai/askai.git
cd askai

# Set OpenAI key (or Azure OpenAI — see .env.example)
cp .env.example .env

# Spin up everything
docker compose up -d

# Ingest a sample corpus (UK Financial Regulation)
make demo-corpus

# Open the chat UI
open http://localhost:3000

That's it. You now have a working knowledge platform with chat, REST API, and MCP server — all running locally, no cloud account needed.

---

Use cases

For companies wanting chat over docs Drop in your SharePoint, Confluence, or S3 documents. Get a chat UI for your team.

For companies building AI agents Use AskAi as the knowledge backbone. Any agent (Claude, LangGraph, CrewAI, custom) can call it via REST or MCP.

For fintech and regulated firms Ingest the regulatory handbooks, supervisory statements, and policy papers your firm has to comply with (FCA, PRA, BoE, equivalents). Build validator agents that check internal reports against the corpus, with paragraph-level citations back to the source rule.

---

Deployment options

Mode	Storage	Vector DB	LLM	Use case
Pure OSS	MinIO	Postgres+pgvector	OpenAI / local LLM	Self-host, no cloud lock-in
Azure	Blob	Azure AI Search	Azure OpenAI	Enterprise on Azure
AWS	S3	OpenSearch / pgvector on RDS	Bedrock / OpenAI	Enterprise on AWS
Hybrid	S3 + Postgres on-prem	pgvector	Azure OpenAI	Mixed regulated environments

All controlled by config. No code changes between modes.

---

Project status

Active development. See Architecture.md for the full design.

---

License

Apache License 2.0 — free for any use, commercial or otherwise, with an explicit patent grant. Built and maintained by FaastLab.Ai.