tandemetry 0.0.1__tar.gz

tandemetry-0.0.1/PKG-INFO

@@ -0,0 +1,93 @@
+Metadata-Version: 2.4
+Name: tandemetry
+Version: 0.0.1
+Summary: Self-hosted analytics for AI coding agents versus human-authored work.
+Keywords: ai-agents,analytics,github,code-review,engineering-metrics
+Author: Yevhenii Korniev
+License-Expression: Apache-2.0
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Software Development
+Requires-Dist: pydantic-settings>=2.6
+Requires-Dist: pyyaml>=6
+Requires-Python: >=3.13
+Project-URL: Homepage, https://github.com/tandemetry/tandemetry
+Project-URL: Repository, https://github.com/tandemetry/tandemetry
+Project-URL: Documentation, https://github.com/tandemetry/tandemetry/tree/main/docs
+Project-URL: Issues, https://github.com/tandemetry/tandemetry/issues
+Description-Content-Type: text/markdown
+
+# Tandemetry
+
+**See how humans and AI agents ship code together.**
+
+Open-source, self-hosted analytics for AI coding agents: compare agent vs human work in your
+repositories — merge rate, reverts, rework, CI failures, review load and cost per PR.
+
+> Status: early development (v0.1 in progress). Not ready for production use.
+
+## Why
+
+AI agents now open a large share of pull requests, but teams can't tell whether that code is
+merged, reverted or rewritten — or which agent works best in their codebase. Tandemetry answers
+that from your own Git data, without sending code anywhere and without LLM calls.
+
+## What you will see
+
+![Dashboard mockup with illustrative data](https://raw.githubusercontent.com/tandemetry/tandemetry/main/docs/assets/dashboard-mockup.svg)
+
+Every number is traceable: each PR is labelled `agent`, `human` or `mixed` by a transparent rule
+(bot author, commit trailer, PR description, branch name), and the matched rule is stored with the
+label. See [agent attribution](docs/attribution.md) and the
+[signatures research](docs/research/signatures.md).
+
+## How it works
+
+GitHub webhooks and a history backfill feed Kafka. The attribution engine labels each event,
+ClickHouse computes the metrics, and Grafana and a metrics API read them. PostgreSQL holds the
+configuration. Details: [architecture](docs/architecture.md).
+
+## Principles
+
+- **Self-hosted.** Your code and data stay in your infrastructure.
+- **Agent-agnostic.** Claude Code, Codex, Copilot coding agent, Devin, Cursor and more.
+- **Outcomes, not activity.** Reverts and rework, not just PR counts.
+- **No individual rankings.** Teams, repositories and agents only.
+- **Deterministic.** Transparent attribution rules, no LLM in the core.
+
+## Docs
+
+- [Product](docs/product.md)
+- [Architecture](docs/architecture.md)
+- [Agent attribution](docs/attribution.md)
+- [Roadmap](docs/roadmap.md)
+- [Releasing](docs/releasing.md)
+- [Architecture decisions](docs/adr/)
+
+## License
+
+[Apache-2.0](LICENSE)
+
+## Development
+
+Requires [uv](https://docs.astral.sh/uv/).
+
+```bash
+uv sync
+uv run pre-commit install
+uv run pytest
+uv run ruff check . && uv run ruff format --check .
+uv run mypy
+uv run tandemetry --version
+```
+
+Local service stack (Kafka, ClickHouse, PostgreSQL):
+
+```bash
+cp deploy/.env.example deploy/.env
+docker compose -f deploy/docker-compose.yml --env-file deploy/.env up -d --wait
+```
+
+See [deploy/README.md](deploy/README.md) for verification and troubleshooting.

tandemetry-0.0.1/README.md

@@ -0,0 +1,72 @@
+# Tandemetry
+
+**See how humans and AI agents ship code together.**
+
+Open-source, self-hosted analytics for AI coding agents: compare agent vs human work in your
+repositories — merge rate, reverts, rework, CI failures, review load and cost per PR.
+
+> Status: early development (v0.1 in progress). Not ready for production use.
+
+## Why
+
+AI agents now open a large share of pull requests, but teams can't tell whether that code is
+merged, reverted or rewritten — or which agent works best in their codebase. Tandemetry answers
+that from your own Git data, without sending code anywhere and without LLM calls.
+
+## What you will see
+
+![Dashboard mockup with illustrative data](https://raw.githubusercontent.com/tandemetry/tandemetry/main/docs/assets/dashboard-mockup.svg)
+
+Every number is traceable: each PR is labelled `agent`, `human` or `mixed` by a transparent rule
+(bot author, commit trailer, PR description, branch name), and the matched rule is stored with the
+label. See [agent attribution](docs/attribution.md) and the
+[signatures research](docs/research/signatures.md).
+
+## How it works
+
+GitHub webhooks and a history backfill feed Kafka. The attribution engine labels each event,
+ClickHouse computes the metrics, and Grafana and a metrics API read them. PostgreSQL holds the
+configuration. Details: [architecture](docs/architecture.md).
+
+## Principles
+
+- **Self-hosted.** Your code and data stay in your infrastructure.
+- **Agent-agnostic.** Claude Code, Codex, Copilot coding agent, Devin, Cursor and more.
+- **Outcomes, not activity.** Reverts and rework, not just PR counts.
+- **No individual rankings.** Teams, repositories and agents only.
+- **Deterministic.** Transparent attribution rules, no LLM in the core.
+
+## Docs
+
+- [Product](docs/product.md)
+- [Architecture](docs/architecture.md)
+- [Agent attribution](docs/attribution.md)
+- [Roadmap](docs/roadmap.md)
+- [Releasing](docs/releasing.md)
+- [Architecture decisions](docs/adr/)
+
+## License
+
+[Apache-2.0](LICENSE)
+
+## Development
+
+Requires [uv](https://docs.astral.sh/uv/).
+
+```bash
+uv sync
+uv run pre-commit install
+uv run pytest
+uv run ruff check . && uv run ruff format --check .
+uv run mypy
+uv run tandemetry --version
+```
+
+Local service stack (Kafka, ClickHouse, PostgreSQL):
+
+```bash
+cp deploy/.env.example deploy/.env
+docker compose -f deploy/docker-compose.yml --env-file deploy/.env up -d --wait
+```
+
+See [deploy/README.md](deploy/README.md) for verification and troubleshooting.

tandemetry-0.0.1/pyproject.toml

@@ -0,0 +1,83 @@
+[project]
+name = "tandemetry"
+version = "0.0.1"
+description = "Self-hosted analytics for AI coding agents versus human-authored work."
+readme = "README.md"
+license = "Apache-2.0"
+keywords = [
+    "ai-agents",
+    "analytics",
+    "github",
+    "code-review",
+    "engineering-metrics",
+]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Developers",
+    "Operating System :: OS Independent",
+    "Programming Language :: Python :: 3.13",
+    "Topic :: Software Development",
+]
+requires-python = ">=3.13"
+dependencies = [
+    "pydantic-settings>=2.6",
+    "pyyaml>=6",
+]
+
+[[project.authors]]
+name = "Yevhenii Korniev"
+
+[project.urls]
+Homepage = "https://github.com/tandemetry/tandemetry"
+Repository = "https://github.com/tandemetry/tandemetry"
+Documentation = "https://github.com/tandemetry/tandemetry/tree/main/docs"
+Issues = "https://github.com/tandemetry/tandemetry/issues"
+
+[project.scripts]
+tandemetry = "tandemetry.cli:main"
+
+[build-system]
+requires = ["uv_build>=0.9.4,<0.10"]
+build-backend = "uv_build"
+
+[dependency-groups]
+dev = [
+    "mypy>=1.13",
+    "pre-commit>=4.0",
+    "pytest>=8.3",
+    "ruff>=0.8",
+    "types-pyyaml>=6.0.12.20261103",
+]
+
+[tool.ruff]
+line-length = 100
+target-version = "py313"
+
+[tool.ruff.lint]
+select = [
+    "E",
+    "F",
+    "I",
+    "UP",
+    "B",
+    "SIM",
+    "N",
+    "ANN",
+    "S",
+    "RUF",
+]
+
+[tool.ruff.lint.per-file-ignores]
+"tests/**" = ["S101"]
+
+[tool.mypy]
+strict = true
+files = [
+    "src",
+    "tests",
+]
+plugins = ["pydantic.mypy"]
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+addopts = "-ra --strict-markers"

tandemetry-0.0.1/pyproject.toml.orig

@@ -0,0 +1,61 @@
+[project]
+name = "tandemetry"
+version = "0.0.1"
+description = "Self-hosted analytics for AI coding agents versus human-authored work."
+readme = "README.md"
+license = "Apache-2.0"
+authors = [{ name = "Yevhenii Korniev" }]
+keywords = ["ai-agents", "analytics", "github", "code-review", "engineering-metrics"]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Developers",
+    "Operating System :: OS Independent",
+    "Programming Language :: Python :: 3.13",
+    "Topic :: Software Development",
+]
+requires-python = ">=3.13"
+dependencies = [
+    "pydantic-settings>=2.6",
+    "pyyaml>=6",
+]
+
+[project.urls]
+Homepage = "https://github.com/tandemetry/tandemetry"
+Repository = "https://github.com/tandemetry/tandemetry"
+Documentation = "https://github.com/tandemetry/tandemetry/tree/main/docs"
+Issues = "https://github.com/tandemetry/tandemetry/issues"
+
+[project.scripts]
+tandemetry = "tandemetry.cli:main"
+
+[build-system]
+requires = ["uv_build>=0.9.4,<0.10"]
+build-backend = "uv_build"
+
+[dependency-groups]
+dev = [
+    "mypy>=1.13",
+    "pre-commit>=4.0",
+    "pytest>=8.3",
+    "ruff>=0.8",
+    "types-pyyaml>=6.0.12.20261103",
+]
+
+[tool.ruff]
+line-length = 100
+target-version = "py313"
+
+[tool.ruff.lint]
+select = ["E", "F", "I", "UP", "B", "SIM", "N", "ANN", "S", "RUF"]
+
+[tool.ruff.lint.per-file-ignores]
+"tests/**" = ["S101"]
+
+[tool.mypy]
+strict = true
+files = ["src", "tests"]
+plugins = ["pydantic.mypy"]
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+addopts = "-ra --strict-markers"

tandemetry-0.0.1/src/tandemetry/attribution/catalog.py

@@ -0,0 +1,85 @@
+from collections.abc import Mapping
+from enum import StrEnum
+from pathlib import Path
+from typing import Self
+
+import yaml
+from pydantic import BaseModel, ConfigDict, Field, ValidationError, model_validator
+
+DEFAULT_CATALOG_PATH = Path(__file__).with_name("signatures.yaml")
+
+
+class Confidence(StrEnum):
+    HIGH = "high"
+    MEDIUM = "medium"
+
+
+class SignalKind(StrEnum):
+    BOT_ACCOUNT = "bot_account"
+    COMMIT_TRAILER = "commit_trailer"
+    DESCRIPTION_PHRASE = "description_phrase"
+    BRANCH_PREFIX = "branch_prefix"
+    LABEL = "label"
+
+
+RULE_CONFIDENCE: Mapping[SignalKind, Confidence] = {
+    SignalKind.BOT_ACCOUNT: Confidence.HIGH,
+    SignalKind.COMMIT_TRAILER: Confidence.HIGH,
+    SignalKind.DESCRIPTION_PHRASE: Confidence.MEDIUM,
+    SignalKind.BRANCH_PREFIX: Confidence.MEDIUM,
+    SignalKind.LABEL: Confidence.MEDIUM,
+}
+
+
+class CatalogError(Exception):
+    """Raised when a signatures catalog cannot be read or is invalid."""
+
+
+class Signal(BaseModel):
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    kind: SignalKind
+    value: str = Field(min_length=1)
+    confidence: Confidence
+
+    @model_validator(mode="after")
+    def _confidence_follows_rules_table(self) -> Self:
+        expected = RULE_CONFIDENCE[self.kind]
+        if self.confidence != expected:
+            raise ValueError(f"{self.kind} signals must have {expected} confidence")
+        return self
+
+
+class AgentSignature(BaseModel):
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    name: str = Field(min_length=1)
+    signals: tuple[Signal, ...] = Field(min_length=1)
+
+
+class SignatureCatalog(BaseModel):
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    agents: tuple[AgentSignature, ...] = Field(min_length=1)
+
+    @model_validator(mode="after")
+    def _agent_names_are_unique(self) -> Self:
+        names = [agent.name for agent in self.agents]
+        duplicates = sorted({name for name in names if names.count(name) > 1})
+        if duplicates:
+            raise ValueError(f"duplicate agent names: {', '.join(duplicates)}")
+        return self
+
+
+class SignatureCatalogLoader:
+    """Loads and validates a signatures catalog from a YAML file."""
+
+    def __init__(self, path: Path = DEFAULT_CATALOG_PATH) -> None:
+        self._path = path
+
+    def load(self) -> SignatureCatalog:
+        try:
+            raw = yaml.safe_load(self._path.read_text(encoding="utf-8"))
+            return SignatureCatalog.model_validate(raw)
+        except (OSError, yaml.YAMLError, ValidationError) as error:
+            raise CatalogError(f"invalid signatures catalog {self._path}: {error}") from error

tandemetry-0.0.1/src/tandemetry/attribution/signatures.yaml

@@ -0,0 +1,31 @@
+agents:
+  - name: Claude Code
+    signals:
+      - {kind: commit_trailer, value: "noreply@anthropic.com", confidence: high}
+      - {kind: description_phrase, value: "Generated with Claude Code", confidence: medium}
+      - {kind: branch_prefix, value: "claude/", confidence: medium}
+  - name: GitHub Copilot coding agent
+    signals:
+      - {kind: bot_account, value: "Copilot", confidence: high}
+      - {kind: bot_account, value: "copilot-swe-agent[bot]", confidence: high}
+      - {kind: commit_trailer, value: "198982749+Copilot@users.noreply.github.com", confidence: high}
+      - {kind: commit_trailer, value: "223556219+Copilot@users.noreply.github.com", confidence: high}
+  - name: Devin
+    signals:
+      - {kind: bot_account, value: "devin-ai-integration[bot]", confidence: high}
+  - name: Cursor
+    signals:
+      - {kind: bot_account, value: "cursor[bot]", confidence: high}
+      - {kind: commit_trailer, value: "cursoragent@cursor.com", confidence: high}
+      - {kind: branch_prefix, value: "cursor/", confidence: medium}
+  - name: OpenAI Codex
+    signals:
+      - {kind: commit_trailer, value: "noreply@openai.com", confidence: high}
+      - {kind: commit_trailer, value: "codex@openai.com", confidence: high}
+      - {kind: branch_prefix, value: "codex/", confidence: medium}
+  - name: Google Jules
+    signals:
+      - {kind: bot_account, value: "google-labs-jules[bot]", confidence: high}
+  - name: Open SWE
+    signals:
+      - {kind: commit_trailer, value: "open-swe@users.noreply.github.com", confidence: high}

tandemetry-0.0.1/src/tandemetry/cli.py

@@ -0,0 +1,40 @@
+import argparse
+from collections.abc import Sequence
+from importlib.metadata import version
+from typing import Protocol
+
+from tandemetry.settings import Settings
+
+
+class VersionProvider(Protocol):
+    def get(self) -> str: ...
+
+
+class PackageVersionProvider:
+    def __init__(self, package: str = "tandemetry") -> None:
+        self._package = package
+
+    def get(self) -> str:
+        return version(self._package)
+
+
+class Cli:
+    def __init__(self, version_provider: VersionProvider, settings: Settings) -> None:
+        self._parser = argparse.ArgumentParser(prog=settings.app_name)
+        self._parser.add_argument(
+            "--version",
+            action="version",
+            version=f"{settings.app_name} {version_provider.get()}",
+        )
+
+    def run(self, argv: Sequence[str] | None = None) -> int:
+        try:
+            self._parser.parse_args(argv)
+        except SystemExit as exit_request:
+            return exit_request.code if isinstance(exit_request.code, int) else 0
+        return 0
+
+
+def main() -> None:
+    cli = Cli(PackageVersionProvider(), Settings())
+    raise SystemExit(cli.run())

tandemetry-0.0.1/src/tandemetry/settings.py

@@ -0,0 +1,7 @@
+from pydantic_settings import BaseSettings, SettingsConfigDict
+
+
+class Settings(BaseSettings):
+    model_config = SettingsConfigDict(env_prefix="TANDEMETRY_")
+
+    app_name: str = "tandemetry"