iranti 0.3.20

Python client for Iranti memory infrastructure

Iranti Python Client

Python client for the Iranti REST API.

Installation

pip install iranti

For local development from this repository:

cd clients/python
pip install -e .

Setup

Start the Iranti API server first:

# In your Iranti directory
npm run api

Usage

from iranti import IrantiClient

client = IrantiClient(
    base_url="http://localhost:3001",
    api_key="replace-with-real-iranti-key"
)

# Or use environment variables
# IRANTI_URL=http://localhost:3001
# IRANTI_API_KEY=replace-with-real-iranti-key
client = IrantiClient()

# Check server is running
print(client.health())

# Write a fact
result = client.write(
    entity="researcher/jane_smith",
    key="affiliation",
    value={"institution": "MIT"},
    summary="Affiliated with MIT",
    confidence=85,
    source="OpenAlex",
    agent="my_agent"
)
print(result.action)  # created / updated / rejected / escalated

# Ingest raw text
result = client.ingest(
    entity="researcher/jane_smith",
    content="Dr. Jane Smith has 24 publications and is at MIT CSAIL.",
    source="OpenAlex",
    confidence=80,
    agent="my_agent"
)
print(
    f"Extracted: {result.extracted_candidates}, "
    f"Written: {result.written}, "
    f"Skipped malformed: {result.skipped_malformed}"
)

# Get working memory before a task
brief = client.handshake(
    agent="my_agent",
    task="Research publication history",
    recent_messages=["Starting research..."]
)
print(brief.inferred_task_type)

# Per-turn memory decision (inject only when needed)
turn = client.attend(
    agent_id="my_agent",
    latest_message="What is my favorite snack?",
    current_context="User: What is my favorite snack?\nAssistant:",
    entity_hints=["user/main"]
)
print(turn["shouldInject"], turn["reason"])

# Save task progress so a later handshake can recommend resuming it
brief = client.checkpoint(
    agent_id="my_agent",
    task="Research publication history",
    recent_messages=["Still comparing two source timelines."],
    checkpoint={
        "currentStep": "Review affiliation conflict",
        "nextStep": "Write corrected fact",
        "openRisks": ["Two sources disagree on start year"],
    },
)

if brief.session_recovery and brief.session_recovery.available:
    client.resume_session(
        agent_id="my_agent",
        session_id=brief.session_recovery.session_id
    )

session = client.inspect_session("my_agent")
print(session.has_checkpoint)

sessions = client.list_sessions(operator_state="interrupted", sort="operator")
print([(item.agent_id, item.operator_state) for item in sessions])

# Query facts
result = client.query("researcher/jane_smith", "affiliation")
if result.found:
    print(result.value)

Error Handling

from iranti import IrantiClient, IrantiAuthError, IrantiValidationError, IrantiError

try:
    result = client.write(...)
except IrantiAuthError:
    print("Invalid API key")
except IrantiValidationError as e:
    print(f"Bad input: {e}")
except IrantiError as e:
    print(f"API error: {e}")

Environment Variables

Variable	Description
IRANTI_URL	API server URL (default: http://localhost:3001)
IRANTI_API_KEY	API token (keyId.secret) or legacy shared server key

License

AGPL-3.0-or-later.