astrai

AI inference in your terminal. PII-stripped, compliance-logged, cost-optimized.

pip install astrai
export ASTRAI_API_KEY=your_key_here
astrai chat "explain this Q function"

Why Astrai?

Your team is already using ChatGPT on personal accounts. You know it.

Astrai gives them the same thing — in the terminal, where they work — but every query is PII-stripped, audit-logged, and routed through approved providers.

• PII stripping — proprietary code never hits provider servers raw
• Compliance audit log — exportable CSV for regulators
• EU/GDPR routing — restrict to European providers only
• Cost optimization — smart routing saves 40-60% vs direct API
• Any model — Claude, GPT, Gemini, DeepSeek, Mistral, local models
• On-prem deployment — astrai deploy for air-gapped environments

CLI Usage

One-shot

astrai chat "write a momentum strategy in Python"
astrai chat -m claude-sonnet-4-5-20250929 "review this code for security issues"
astrai chat --privacy enhanced "analyze this proprietary algorithm"
astrai chat --region eu "process this customer data"

Pipe input

cat strategy.q | astrai chat "optimize this Q code for memory usage"
git diff | astrai chat "review these changes"
echo "select from trade where sym=\`AAPL" | astrai chat "explain this query"

Interactive mode

$ astrai chat
astrai interactive mode (ctrl+d to exit)
  model: auto | privacy: standard | region: any

astrai> explain the difference between aj and wj in kdb+

On-prem deployment

pip install astrai[cli]
astrai deploy --mode local --profile fund

Generates Docker Compose files, pulls model weights, and starts a fully local inference stack. No data leaves your network.

Python SDK

from astrai import Astrai

# Drop-in OpenAI replacement
client = Astrai(budget_limit=5.00)

response = client.chat.completions.create(
    messages=[{"role": "user", "content": "Optimize this Q query"}],
    model="auto",  # Astrai picks best model for the task
)

print(response.choices[0].message.content)
print(response.astrai.cost_usd)       # What you paid
print(response.astrai.provider_used)   # Which provider was selected

Agent mode (budget-controlled)

from astrai import AstraiAgent

agent = AstraiAgent(
    agent_id="quant-research",
    budget_limit=10.00,
    region="eu",
)

response = agent.chat("Write a pairs trading strategy")
print(agent.spend)      # $0.003
print(agent.remaining)  # $9.997

Configuration

# Required
export ASTRAI_API_KEY=your_key_here

# Optional
export ASTRAI_BASE_URL=https://as-trai.com/v1
export ASTRAI_MODEL=auto
export ASTRAI_PRIVACY=standard    # standard|enhanced|max
export ASTRAI_REGION=any          # any|eu|us

Privacy Modes

Mode	PII Stripping	Server Logging	Use Case
standard	No	Full	Development, testing
enhanced	Yes	Metadata only	Production, sensitive code
max	Yes	None	Regulated industries, EU

Pricing

• Free: 50 requests/day, standard privacy
• Pro ($49/mo): Unlimited, enhanced privacy, EU routing, PII stripping
• Enterprise ($999/mo): Team accounts, compliance exports, SLA, on-prem support

Get a free key at as-trai.com