clovis 0.5.13

Serveur IA de Clovis — AKA cloclo

clovis

Web Search · Deep Research · RAG · Embeddings · Structured Outputs

---

clovis is a Python client for the cloooooo.com AI API — and optionally a self-hosted server for those running their own GPU stack. It ships with multi-step web research, a full RAG pipeline, vector embeddings, reranking, structured JSON outputs, vision, and an agentic deep-research mode — all accessible via a single HTTP endpoint.

Features

• Simple inference — one-line calls with streaming, negative prompts, and extended reasoning
• Web search — live results injected into context, always date-aware
• Deep thinking — multi-step agentic research pipeline with source citations
• Ultra deep thinking — multi-axis research with automated gap analysis, 280+ sources synthesized into a structured report
• RAG — ingest PDF, DOCX, TXT documents; semantic search over your corpus
• Embeddings — 768-dim dense vectors
• Reranking — cross-encoder reranking of document candidates
• Structured output — JSON Schema-constrained generation
• Vision — image description from URL, file path, or base64
• Auto-routing — automatic mode selection based on query type
• Conversation memory — short-term history per conversation ID

---

Installation

pip install clovis

Requires Python 3.10+

---

Quick start

No local setup required — connect directly to the hosted server:

from clovis import cloooooo

# Option A — Hosted server (no GPU, no config)
ai = cloooooo(base_url="http://cloooooo.com")
print(ai("Explain transformer architecture"))

# Option B — Self-hosted (see "Self-hosting" section below)
ai = cloooooo()  # connects to localhost:61005

# With options
response = ai(
    "Write a sonnet about entropy",
    negative_prompt="no rhymes",
    thinking=True,
    context="You are a physicist who loves poetry.",
)

# Streaming
for token in ai.stream("Describe the Big Bang in detail"):
    print(token, end="", flush=True)

# Multi-turn conversation
conv = ai.conversation(context="You are a senior software engineer.")
conv("Explain dependency injection")
conv("Show me a Python example")
conv("How would you test it?")

---

API server

All endpoints accept Content-Type: application/json. Streaming responses use text/plain.

Base URL: http://cloooooo.com (hosted) or http://localhost:8000 (self-hosted).

---

POST /ia — Universal endpoint

The main endpoint. Handles all inference modes.

curl -X POST http://cloooooo.com/ia \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What is quantum entanglement?", "use_web": true}'

Parameters

Parameter	Type	Default	Description
prompt	str	required	The question or instruction
mode	str	null	"simple" · "deep_thinking" · "ultra_deep_thinking"
use_web	bool	false	Inject live web search results with current date
thinking	bool	false	Enable extended reasoning (chain-of-thought)
stream	bool	false	Stream tokens via text/plain
use_memory	bool	false	Load and save conversation history
conversation_id	str	null	Key for conversation memory
context	str	null	System-level context injected before the prompt
negative_prompt	str	null	Instructions for what to avoid

Response

{"response": "Quantum entanglement is a phenomenon where..."}

For deep_thinking:

{
  "answer": "...",
  "sources": ["https://...", "https://..."],
  "model_used": "miroflow:Qwen/Qwen3-32B-AWQ",
  "fallback_used": false
}

For ultra_deep_thinking (async job):

{"job_id": "b45d79...", "status": "pending", "poll": "/job/b45d79..."}

---

Modes

simple — Direct inference

Fast, direct LLM call. Optionally augmented with web search or reasoning.

import httpx

r = httpx.post("http://cloooooo.com/ia", json={
    "prompt": "Latest news on fusion energy",
    "use_web": True,
    "thinking": True,
})
print(r.json()["response"])

---

deep_thinking — Agentic web research

Multi-step research pipeline. Performs web searches, reasons over the results, and returns a structured answer with source citations.

r = httpx.post("http://cloooooo.com/ia", json={
    "prompt": "What are the geopolitical implications of AGI development?",
    "mode": "deep_thinking",
}, timeout=300)

data = r.json()
print(data["answer"])
print(data["sources"])

Streaming mode returns progress updates then the final JSON:

curl -N -X POST http://cloooooo.com/ia \
  -d '{"prompt": "Impact of interest rates on tech stocks", "mode": "deep_thinking", "stream": true}'

# [deep_thinking... 5s]
# [deep_thinking... 10s]
# ...
# {"answer": "...", "sources": [...], "fallback_used": false}

---

ultra_deep_thinking — Multi-axis deep research

The most thorough mode. Decomposes the question into research axes, searches each independently, identifies knowledge gaps, fills them, then synthesizes a structured report. Typically produces 10 000–15 000 character reports with 250–300 unique sources. Runs as an async job (15–25 min).

import httpx, time

# 1. Submit
r = httpx.post("http://cloooooo.com/ia", json={
    "prompt": "How does reinforcement learning from human feedback (RLHF) work?",
    "mode": "ultra_deep_thinking",
})
job_id = r.json()["job_id"]

# 2. Poll until done
while True:
    job = httpx.get(f"http://cloooooo.com/job/{job_id}").json()
    print(f"status: {job['status']} | {len(job['progress'])} steps logged")
    if job["status"] == "done":
        print(job["result"]["answer"][:500])
        break
    time.sleep(30)

Progress visible at each poll:

[décomposition] 5 axes : ['Définition', 'Historique', ...]
[axe:Définition] recherche en cours...
[axe:Définition] OK — 4 127 chars, 36 sources
...
[lacunes 1/2] 5 lacunes identifiées → 5/5 comblées
[synthèse] 15 sections · 65 000 chars · 295 sources...
[terminé] 14 067 chars · 280 sources uniques

Presets (via /ultra_deep_thinking endpoint):

Preset	Axes	Gap rounds	Duration
fast	3	1	~5 min
deep (default)	5	2	~15 min
ultra	8	3	~30 min

---

GET /health — Server status

curl http://cloooooo.com/health

{
  "status": "ok",
  "version": "0.5.12",
  "model": "Qwen/Qwen3-32B-AWQ",
  "modes": ["simple", "search", "thinking", "deep_thinking", "ultra_deep_thinking", "embed", "rerank", "vision"]
}

---

POST /embed — Text embeddings

r = httpx.post("http://cloooooo.com/embed", json={
    "texts": ["Hello world", "Machine learning basics"],
})
print(r.json()["dim"])          # 768
print(len(r.json()["embeddings"]))  # 2

---

POST /rerank — Document reranking

r = httpx.post("http://cloooooo.com/rerank", json={
    "query": "machine learning optimization",
    "documents": [
        "Gradient descent is an optimization algorithm for ML",
        "The weather in Paris is sunny today",
        "Adam optimizer adapts learning rates per parameter",
    ],
    "top_k": 2,
})
for item in r.json()["results"]:
    print(f"{item['score']:.3f}  {item['document'][:60]}")

---

POST /structured — JSON Schema output

r = httpx.post("http://cloooooo.com/structured", json={
    "prompt": "Describe the movie Inception",
    "schema": {
        "type": "object",
        "properties": {
            "title": {"type": "string"},
            "year": {"type": "integer"},
            "genres": {"type": "array", "items": {"type": "string"}},
        },
        "required": ["title", "year", "genres"],
    },
})
print(r.json()["result"])
# {"title": "Inception", "year": 2010, "genres": ["sci-fi", "thriller"]}

---

POST /vision — Image understanding

r = httpx.post("http://cloooooo.com/vision", json={
    "image": "https://example.com/photo.jpg",
    "prompt": "What objects do you see?",
})
print(r.json()["response"])

---

POST /rag/ingest + POST /rag/ask — RAG

httpx.post("http://cloooooo.com/rag/ingest", json={"path": "/path/to/doc.pdf"})

r = httpx.post("http://cloooooo.com/rag/ask", json={
    "question": "What are the main conclusions?",
    "top_k": 5,
})
print(r.json()["response"])

Supported formats: PDF, DOCX, TXT, Markdown.

---

Other endpoints

Endpoint	Method	Description
/job/{id}	GET	Poll async job status + progress
/route	POST	Auto-select the best mode for a prompt
/deep_think	POST	Standalone iterative deep research (streaming)
/tools	GET	List available tools
/rag/sources	GET	List ingested documents
/openapi.json	GET	OpenAPI schema
/docs	GET	Interactive API documentation

---

Streaming

import httpx

with httpx.stream("POST", "http://cloooooo.com/ia", json={
    "prompt": "Write a detailed explanation of CRISPR-Cas9",
    "stream": True,
}) as r:
    for chunk in r.iter_text():
        print(chunk, end="", flush=True)

---

Self-hosting

These components run server-side on cloooooo.com. No local installation is required if you use the hosted server.

To run your own instance (requires an NVIDIA GPU with 24 GB+ VRAM):

# 1. Start SGLang with Qwen3-32B-AWQ
python -m sglang.launch_server \
  --model-path Qwen/Qwen3-32B-AWQ \
  --port 61005 \
  --quantization awq

# 2. Start SearXNG (for web search modes)
docker run -p 8888:8080 searxng/searxng

# 3. Start clovis
clovis serve --port 8000

# Environment variables
export CLOVIS_LOCAL_URL="http://localhost:61005"
export CLOVIS_MODEL="Qwen/Qwen3-32B-AWQ"
export CLOVIS_API_KEY="sk-..."
export SEARXNG_URL="http://localhost:8888"

Then point the client to your server:

ai = cloooooo(base_url="http://your-server:8000")

---

License

MIT — Clovis Sfeir