Skip to main content
Caveman distributes its developer surface as standalone packages for application integration, schemas, evaluation, and framework adapters. This page explains how the packages fit together, what each one covers, and how to build and test them.

​
TypeScript and Python SDKs

The TypeScript SDK lives in packages/sdk/typescript and the Python SDK lives in packages/sdk/python. Both implement the same high-level operations:
  • Create a Caveman client
  • Configure provider calls
  • Define, defer, and search tools
  • Compress eligible context
  • Assemble a request from context parts
  • Create and consume context packs
  • Emit traces and OpenTelemetry data
  • Apply runtime policy
Field names and /sdk/v1/* request contracts stay aligned across both languages. A contract change is incomplete until implementations, schemas, and tests agree in both SDKs. Install the TypeScript SDK:
Install the Python SDK:
Build and test the TypeScript package:
Test the Python package:
Neither SDK guesses cost for an unknown model. Unknown pricing remains zero and is explicitly marked unpriced.

​
Shared contracts

packages/shared/contracts stores JSON Schema wire contracts. The current schema set covers:
  • Adapter conformance and agent-run receipts
  • Cache guards and canonical spans
  • Cave builds and Cave Plans
  • Context intermediate representation
  • Continuous-improvement reports
  • Evaluation cases and grader registry
  • Harness events and policy
  • Practices
  • Transform capabilities and traces
Generate and validate artifacts through package scripts rather than editing generated outputs independently.

​
Provider catalog

shared/provider-catalog stores dated public list-price records and generated catalog outputs for local estimates. An unsupported model returns zero price plus an unpriced marker. The catalog does not represent invoice data. Catalog updates need source date, provider unit semantics, generated-artifact refresh, and tests. See Accounting and evidence for how list-price subtotals relate to real billing.

​
Benchmark tooling

packages/subagent-tax measures local context and delegation fixtures without issuing provider requests. Its output is benchmark evidence for exact fixtures and counterfactual implementation, not a general savings claim.

​
Package release model

Registry packages release independently through scoped workflow inputs, while native binaries use a separate signed process. See PACKAGE_RELEASES.md and Installation for details.