L'Intello

The brainy one — a smart AI backend for multi-LLM routing, literary analysis, writing tools, OCR, and document reconstruction.

git clone https://github.com/collaed/lintello.git
cd intello
cp .env.example .env    # Add your API keys
docker compose up -d    # → http://localhost:8000

Get a free Groq key at https://console.groq.com — that's enough to start.

29 models across 13 providers. Ask a question → L'Intello picks the best free model, falls back on failure, caches responses, and learns from your feedback.

Upload a novel (TXT/PDF/EPUB) → get chapter structure, character tracking (spaCy NER), pacing curves, narrative thread visualization, and AI-powered edit suggestions.

Show-not-tell, 5-sense describe, tone shift, brainstorm, shrink ray, first draft generator, 3 AI beta readers in parallel.

Project briefs → adaptive next-step → horizontal (expand) / vertical (enrich) modes → word count tracking → resumable across sessions.

Tesseract → OCR.space → Gemini Vision auto-escalation. Single images, PDFs, async jobs for large books. 9 languages.

Upload 50+ scattered version files → detect cross-references → rebuild complete document → LLM-smooth transitions.

Drop-in replacement at /v1/chat/completions. Works with any OpenAI SDK client, Ollama, external clients.

intello/
├── web.py              FastAPI app — 70+ routes, auth, all endpoints
├── backends.py         LLM execution (14 providers)
├── router.py           Task classification + scoring
├── pipeline.py         Deep mode (draft → review → synthesis)
├── debate.py           Multi-model adversarial debate
├── chains.py           Prompt chaining / task decomposition
├── literary.py         Document ingestion, structure, pacing, edits
├── workflow.py         Writing workflow engine
├── writing_tools.py    AI writing transformations
├── craft.py            Dynamic literary reference engine
├── reconstruct.py      Version reconstruction from scattered files
├── nlp.py              spaCy NER + linguistic analysis
├── cache.py            Semantic cache (sentence-transformers)
├── memory.py           Conversation memory + learning
├── guardrails.py       Anti-hallucination + word count
├── tools.py            Web search, calculator, Python eval
├── ocr.py              Tesseract + OCRmyPDF
├── ocr_engines.py      Multi-engine OCR escalation
├── imagegen.py         Image generation routing
├── scheduler.py        Recurring tasks
├── webhooks.py         External integrations
├── gdrive.py           Google Drive OAuth + browsing
├── keys.py             API key management
├── ratelimit.py        Daily quota tracking
├── models.py           Data models
├── research.py         Provider catalog (29 models)
└── static/
    ├── index.html      Chat UI (ChatGPT-style)
    ├── literary.html   Literary analysis page
    ├── corkboard.html  Visual scene board
    └── gdrive.html     Google Drive file browser

• 6,100+ lines of Python
• 1,500+ lines of HTML/JS
• 70+ API routes
• 29 LLM providers (20 free, 41,850 free requests/day)
• 36 automated tests (whitebox + greybox + blackbox)

L'Intello was built as a personal tool — a smart AI backend that routes prompts to the best free model, analyzes manuscripts, OCRs documents, and reconstructs scattered version files. It turned out to be useful enough to share.

Most AI tools lock you into one provider, charge per word, or trap your data. L'Intello does the opposite:

• Bring your own keys — use free tiers from 13 providers, pay nothing
• Self-hosted — your data stays on your server
• No vendor lock-in — swap providers by editing one file
• No subscription — 48,000+ free requests/day across 26 free models

It's not polished commercial software. It's a working tool built by a writer who got tired of paying $59/month for commercial AI writing tools when free LLMs are this good.

AGPL-3.0 — free to use, modify, and self-host. If you modify and run it as a network service, share your changes.