kurrent-agent-schema
Canonical event schema types for Kurrent agent integrations — schema version 2.
This package is the Python mirror of the canonical agent event schema shared across Kurrent's agent-framework integrations (Google ADK, Microsoft Agent Framework, Strands, OpenAI Agents, Claude Agent SDK) and Capacitor. The prose specification lives in schema/SCHEMA_v2.md; the .NET mirror is Kurrent.Agent.Schema.
Wire format
Backed by Protobuf codegen from schema/proto/. The JSON wire format follows the proto3 canonical mapping with preserve_proto_field_name (snake_case on the wire). The sanctioned JSON entry points are to_json(event) and from_json(cls, src) — direct calls to google.protobuf.json_format are not supported.
What's here
- Canonical event models (generated Protobuf message classes):
SessionStarted,SessionEnded,SessionContinuedAs,UserMessageReceived,AssistantTextGenerated,AssistantToolCallsGenerated,AssistantThinkingGenerated,ToolResultReceived,InterruptIssued,InterruptResolved,SubagentStarted,SubagentCompleted,FactRetained,ArtifactVersionCreated,EvalRunStarted,TurnScored,EvalRunCompleted. - Value types:
AgentConfig,ToolSpec,ToolCallInfo. - Usage metadata:
TokenUsage(carried on KurrentDB event metadata under the$usagekey, not as a standalone event). - Stream-name builders:
agent_session_stream,agent_subsession_stream,agent_memory_stream,agent_artifact_stream,eval_run_stream. - Type-name registry:
EVENT_TYPE_NAMES/EVENT_TYPE_BY_NAMEfor wiring up KurrentDB event-type to/from CLR-type conversion.
What's not here
- Framework-specific events and extension shapes. ADK's
AgentTransferred/Rewind, AFW's workflow checkpoints, Capacitor'sAgentRunStarted/visibility events, coding-agent fields underextensions.claude_code.*, etc. Those live in their owning integration packages — this package deliberately carries only the portable vocabulary.
Usage
from datetime import datetime, timezone
from kurrent_agent_schema import (
SessionStarted, UserMessageReceived, AssistantTextGenerated,
TokenUsage, USAGE_METADATA_KEY, agent_session_stream,
)
stream = agent_session_stream("sess-0001") # "AgentSession-sess-0001"
event = UserMessageReceived(
content="hello",
message_index=0,
timestamp=datetime.now(tz=timezone.utc),
extensions={"adk": {"invocation_id": "inv_abc"}},
)
# $usage goes on KurrentDB metadata, not the payload
usage = TokenUsage(input_tokens=1507, output_tokens=203, model="claude-sonnet-4-6")
Drift guard
Every canonical event has a JSON fixture under schema/fixtures/events/. The test suite here runs a round-trip assertion per fixture; an equivalent suite in the .NET package runs against the same fixtures. Adding or modifying a canonical field requires a coordinated PR: update the model and the fixture, in both Python and .NET packages. CI fails otherwise.
cd schema/python
uv pip install -e ".[dev]"
pytest
Version
- Package:
0.5.0(Protobuf-generated source; seeCHANGELOG.mdfor breaking changes). - Schema:
SCHEMA_VERSION = 2, stamped on KurrentDB metadata under$schema_versionby integration writers.
Metadata
Release files for kurrent-agent-schema 0.5.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| kurrent_agent_schema-0.5.0.tar.gz | 20.7 kB | With provenance | Details for kurrent_agent_schema-0.5.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| kurrent_agent_schema-0.5.0-py3-none-any.whl | Python 3 | none | any | Details for kurrent_agent_schema-0.5.0-py3-none-any.whl |
Total release size: 37.9 kB
Release files / kurrent_agent_schema-0.5.0.tar.gz
| Download URL | kurrent_agent_schema-0.5.0.tar.gz |
|---|---|
| Size | 20.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
a58d644151f3e66633102c569fdd76cb34d6e9a93de67ca66df61c8847499c48
|
|
BLAKE2b-256 checksum How to use checksums |
1bda88972bb9a1b9b242c47c5ed08c5f6428c2ac4432ec0730190a7eaa00ac36
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 7, 2026.
Transparency logRelease files / kurrent_agent_schema-0.5.0-py3-none-any.whl
| Download URL | kurrent_agent_schema-0.5.0-py3-none-any.whl |
|---|---|
| Size | 17.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
8e8f66950edd7fcb201a35a18a3b31f57630ed5ffd3024e92ecc20779b7f1f4f
|
|
BLAKE2b-256 checksum How to use checksums |
5acb963df41a672cb33ce7a0af64188490bf2983e6322a05b877bc18ac42f223
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 7, 2026.
Transparency logSecurity for kurrent-agent-schema 0.5.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for kurrent_agent_schema-0.5.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 7, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.