Skip to content

v0.12.0 — TealTiger Governance Integration

Choose a tag to compare

View all tags
@ferhimedamine ferhimedamine released this 15 Jun 04:48
· 27 commits to main since this release
v0.12.0
147a45c

What's New in v0.12.0

TealTiger Governance Integration

First-class integration with the TealTiger governance framework. Three new adapter classes let you use Dakera as the persistent memory backend for TealTiger's cost tracking, governance decision receipts, and delegation chains.

Install:

pip install dakera[tealtiger]

New in dakera.integrations.tealtiger:

DakeraCostStorage

Implements TealTiger's CostStorage ABC with all 8 async methods. Persists cost records as Dakera memories, enabling LLM cost lifecycle tracking with importance-based retention.

from dakera import AsyncDakeraClient
from dakera.integrations.tealtiger import DakeraCostStorage

client = AsyncDakeraClient(api_key="...", base_url="...")
cost_storage = DakeraCostStorage(client, agent_id="my-agent")

DakeraDecisionStore

Receipt persistence for TealTiger governance decisions. Maps Decision objects to Dakera memories with action-type-derived importance. is_terminal() correctly distinguishes pending decisions (REQUIRE_APPROVAL) from resolved ones (ALLOW/DENY/TIMED_OUT).

from dakera.integrations.tealtiger import DakeraDecisionStore

decision_store = DakeraDecisionStore(client, agent_id="my-agent")
await decision_store.store_receipt(decision)
terminal = await decision_store.is_terminal(decision_id)

DakeraDelegationHelper

Knowledge-graph-backed delegation chain tracker. Stores delegation relationships as linked Dakera memories for full graph traversal of agent hierarchies.

from dakera.integrations.tealtiger import DakeraDelegationHelper

delegation = DakeraDelegationHelper(client, agent_id="my-agent")
await delegation.link_delegation(delegator_id, delegate_id, context)
chain = await delegation.get_delegation_chain(delegate_id)

See examples/tealtiger_governance.py for a complete end-to-end example.

Special thanks to @nagasatish007 (TealTiger maintainer) for reviewing this integration and providing valuable feedback that improved the is_terminal() semantics and error handling.

TifScore — Truth-Indeterminacy-Falsity Reliability Scoring

New TifScore dataclass in dakera.models for T-I-F reliability scoring of stored memories.

from dakera import DakeraClient
from dakera.models import TifScore

client = DakeraClient(api_key="...", base_url="...")
score = client.evaluate_tif(memory_id)
print(score.classification)  # "confident_reuse" | "ask_clarification" | ...

Coverage

  • 34 new tests for TealTiger adapters (with/without tealtiger installed)
  • 1 new example — examples/tealtiger_governance.py
  • All 429 tests pass on Python 3.10 and 3.12

Full Changelog: https://github.com/Dakera-AI/dakera-py/blob/main/CHANGELOG.md#0120---2026-06-15

Contributors

nagasatish007
Assets 2
Loading