Pramiti Docs

Standalone Packages

Open-source Python packages that work independently of the Pramiti platform

Pramiti publishes four standalone Python packages that solve specific problems in the AI agent ecosystem. Each package is zero-dependency on the Pramiti platform and can be installed and used independently.

Packages

PackageInstallPurpose
pramiti-agent-finopspip install pramiti-agent-finopsPer-agent cost caps and attribution
pramiti-export-copilotpip install pramiti-export-copilotPush semantic models to dbt, Snowflake, Looker, Tableau
pramiti-merger-reconcilerpip install pramiti-merger-reconcilerCross-ontology class matching for M&A

The Flight Recorder package is documented separately at Flight Recorder.

Design Philosophy

All packages follow these principles:

  1. Zero platform dependency — Each package works standalone with no Pramiti platform imports
  2. Pluggable backends — Storage and delivery targets are injected via interfaces, not hardcoded
  3. Minimal external dependencies — Core logic uses only stdlib; optional integrations are extras
  4. Works with or without Pramiti — When running inside the platform, packages use platform backends; when standalone, they use in-memory or file-based backends

Platform Integration

When used within the Pramiti platform, packages use platform-specific backends:

# Standalone
from pramiti_agent_finops import CapChecker
from pramiti_agent_finops.backends import InMemoryBackend
checker = CapChecker(InMemoryBackend())
 
# Platform-integrated
from pramiti_agent_finops import CapChecker
from epistom.services.finops import PostgreSQLCostBackend
checker = CapChecker(PostgreSQLCostBackend(db_session))

On this page