The infrastructure layer that makes AI trustworthy.

Every AI system has three unsolved problems: it costs too much, it gets things wrong, and nobody can prove it behaved correctly. Ombre solves all three — running entirely inside your own infrastructure.

Your data never leaves your environment. You bring your own API keys.

# Base install
pip install git+https://github.com/ombre-ai/ombre-core.git

# With your provider
pip install "git+https://github.com/ombre-ai/ombre-core.git#egg=ombre-ai[openai]"
pip install "git+https://github.com/ombre-ai/ombre-core.git#egg=ombre-ai[anthropic]"
pip install "git+https://github.com/ombre-ai/ombre-core.git#egg=ombre-ai[groq]"
pip install "git+https://github.com/ombre-ai/ombre-core.git#egg=ombre-ai[all-providers]"

git clone https://github.com/ombre-ai/ombre-core.git
pip install ./ombre-core

from ombre import Ombre

ai = Ombre(
    openai_key="sk-...",        # Your key — Ombre never sees it
    # anthropic_key="sk-ant-...",
    # groq_key="gsk-...",
)

response = ai.run("Summarize our Q3 financials and recommend next steps")

print(response.text)                  # The answer
print(response.confidence)            # Verified confidence score (0.0–1.0)
print(response.cost_saved)            # Dollars saved vs raw API call
print(response.audit_id)              # Immutable audit reference
print(response.hallucinations_caught) # Bad answers stopped before reaching you
print(response.threats_blocked)       # Security events intercepted

response = ai.chat([
    {"role": "user", "content": "My name is Alex. I work in finance."},
    {"role": "assistant", "content": "Got it Alex, how can I help?"},
    {"role": "user", "content": "What should I focus on this quarter?"},
])
# Memory Agent remembers Alex works in finance — no need to repeat it

responses = ai.batch([
    "Summarize contract 1",
    "Summarize contract 2",
    "Summarize contract 3",
], concurrency=5)

# Verified facts — model uses these instead of guessing
ai.truth.add_fact(
    key="company_ceo",
    fact="The CEO of Acme Corp is Jane Smith, appointed January 2022",
    confidence=1.0,
    source="company_records",
)

# Export for compliance
ai.export_audit("audit.json", format="json")
ai.export_audit("audit.csv", format="csv")

# EU AI Act compliance report
ai.audit.generate_compliance_report(
    regulation="eu_ai_act",
    output_path="compliance_report.json",
)

# Start the server
python -m ombre serve --port 8080

# Or from CLI
ombre serve --port 8080

# cURL
curl -X POST http://localhost:8080/v1/run \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Your prompt here"}'

# Python
import requests
r = requests.post("http://localhost:8080/v1/run",
    json={"prompt": "Your prompt here"})
print(r.json()["text"])

// JavaScript / Node
const res = await fetch("http://localhost:8080/v1/run", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ prompt: "Your prompt here" })
})
const data = await res.json()
console.log(data.text)

// Go
resp, _ := http.Post("http://localhost:8080/v1/run",
    "application/json",
    strings.NewReader(`{"prompt":"Your prompt here"}`))

export OPENAI_API_KEY=sk-...
export ANTHROPIC_API_KEY=sk-ant-...
export GROQ_API_KEY=gsk-...
export OMBRE_API_KEY=omb_ent_...   # Enterprise license key (optional)

ai = Ombre()  # Reads keys from environment automatically

Your Application
      │
      ▼
┌─────────────────────────────────────┐
│           OMBRE LAYER               │
│  (runs inside your infrastructure)  │
│                                     │
│  1. Security Agent                  │
│  2. Memory Agent                    │
│  3. Token Agent  ◄── cache hit?     │
│  4. Compute Agent                   │
│  5. Truth Agent                     │
│  ── model inference ──              │
│  6. Latency Agent                   │
│  7. Reliability Agent               │
│  8. Audit Agent                     │
└─────────────────────────────────────┘
      │
      ▼
Your AI Model
(OpenAI / Anthropic / Groq / Mistral)