agentforge-otel 0.2.3__tar.gz

agentforge_otel-0.2.3/.gitignore

@@ -0,0 +1,49 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+
+# Distribution / packaging
+.Python
+build/
+dist/
+*.egg-info/
+*.egg
+wheels/
+sdist/
+share/python-wheels/
+
+# uv
+.venv/
+uv-cache/
+
+# Testing
+.pytest_cache/
+.coverage
+.coverage.*
+htmlcov/
+coverage.xml
+.tox/
+.mypy_cache/
+.ruff_cache/
+.hypothesis/
+
+# Environment
+.env
+.envrc
+
+# Editors
+.vscode/
+.idea/
+*.swp
+*.swo
+*~
+
+# OS
+.DS_Store
+Thumbs.db
+
+# Project-local
+*.local
+.agentforge-state/.session-cache

agentforge_otel-0.2.3/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

agentforge_otel-0.2.3/PKG-INFO

@@ -0,0 +1,90 @@
+Metadata-Version: 2.4
+Name: agentforge-otel
+Version: 0.2.3
+Summary: OpenTelemetry tracing + metrics emitter for AgentForge
+Project-URL: Homepage, https://github.com/Scaffoldic/agentforge-py
+Project-URL: Repository, https://github.com/Scaffoldic/agentforge-py
+Project-URL: Documentation, https://github.com/Scaffoldic/agentforge-py
+Project-URL: Changelog, https://github.com/Scaffoldic/agentforge-py/blob/main/CHANGELOG.md
+Project-URL: Issues, https://github.com/Scaffoldic/agentforge-py/issues
+Author: The AgentForge Authors
+License-Expression: Apache-2.0
+License-File: LICENSE
+Keywords: agent,ai,observability,opentelemetry,tracing
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: System :: Monitoring
+Classifier: Typing :: Typed
+Requires-Python: >=3.13
+Requires-Dist: agentforge-core~=0.2.3
+Requires-Dist: opentelemetry-exporter-otlp-proto-grpc>=1.27
+Requires-Dist: opentelemetry-sdk>=1.27
+Description-Content-Type: text/markdown
+
+# agentforge-otel
+
+OpenTelemetry tracing for AgentForge (feat-009).
+
+The OTel **API** ships in `agentforge-core` — the framework emits
+spans unconditionally. Without this package, those calls degrade to
+`NonRecordingSpan` (near-zero cost). Installing `agentforge-otel`
+wires up the **SDK**, an OTLP exporter, and a sampler so the spans
+actually reach a collector.
+
+## Quick start
+
+```python
+from agentforge import Agent
+from agentforge_otel import OpenTelemetryHook
+
+otel = OpenTelemetryHook(
+    endpoint="http://otel-collector:4317",
+    service_name="my-agent",
+    sample_rate=1.0,
+)
+
+agent = Agent(
+    model="bedrock:...",
+    tools=[...],
+    on_step=otel,
+    on_finish=otel,
+)
+```
+
+Constructing the hook installs the SDK provider once (idempotent). The
+framework's existing span emission then produces a tree per run:
+
+    span: agent.run
+    └── span: strategy.iteration
+        ├── span: llm.call
+        ├── span: tool.<name>
+        └── span: evaluator.<name>
+
+Span attributes carry `agentforge.run_id`, cost, token counts,
+finish reason, and step count — same correlation key the
+`RunIdFilter` puts on every log line.
+
+## Multiple backends
+
+Run OTel alongside any other observer:
+
+```python
+agent = Agent(
+    model="...",
+    on_step=[otel, my_statsd_hook],
+    on_finish=[otel, persist_to_db],
+)
+```
+
+Hooks fire in registration order; one hook's exception doesn't
+affect siblings (the framework logs WARN via
+`agentforge.observability` and keeps going).
+
+## Vendor compatibility
+
+OTel is the wire format; any collector that ingests OTLP works:
+Datadog, Honeycomb, Jaeger, Grafana Tempo, SigNoz, New Relic. Point
+the `endpoint` at your collector's gRPC port (4317 by default).

agentforge_otel-0.2.3/README.md

@@ -0,0 +1,64 @@
+# agentforge-otel
+
+OpenTelemetry tracing for AgentForge (feat-009).
+
+The OTel **API** ships in `agentforge-core` — the framework emits
+spans unconditionally. Without this package, those calls degrade to
+`NonRecordingSpan` (near-zero cost). Installing `agentforge-otel`
+wires up the **SDK**, an OTLP exporter, and a sampler so the spans
+actually reach a collector.
+
+## Quick start
+
+```python
+from agentforge import Agent
+from agentforge_otel import OpenTelemetryHook
+
+otel = OpenTelemetryHook(
+    endpoint="http://otel-collector:4317",
+    service_name="my-agent",
+    sample_rate=1.0,
+)
+
+agent = Agent(
+    model="bedrock:...",
+    tools=[...],
+    on_step=otel,
+    on_finish=otel,
+)
+```
+
+Constructing the hook installs the SDK provider once (idempotent). The
+framework's existing span emission then produces a tree per run:
+
+    span: agent.run
+    └── span: strategy.iteration
+        ├── span: llm.call
+        ├── span: tool.<name>
+        └── span: evaluator.<name>
+
+Span attributes carry `agentforge.run_id`, cost, token counts,
+finish reason, and step count — same correlation key the
+`RunIdFilter` puts on every log line.
+
+## Multiple backends
+
+Run OTel alongside any other observer:
+
+```python
+agent = Agent(
+    model="...",
+    on_step=[otel, my_statsd_hook],
+    on_finish=[otel, persist_to_db],
+)
+```
+
+Hooks fire in registration order; one hook's exception doesn't
+affect siblings (the framework logs WARN via
+`agentforge.observability` and keeps going).
+
+## Vendor compatibility
+
+OTel is the wire format; any collector that ingests OTLP works:
+Datadog, Honeycomb, Jaeger, Grafana Tempo, SigNoz, New Relic. Point
+the `endpoint` at your collector's gRPC port (4317 by default).

agentforge_otel-0.2.3/pyproject.toml

@@ -0,0 +1,62 @@
+# agentforge-otel — OpenTelemetry tracing + metrics for AgentForge.
+#
+# Tier-3 module per ADR-0003. Installing this package wires up the
+# OTel SDK (provider + OTLP exporter + sampler) so that the framework
+# spans emitted from `agentforge-core/observability/tracing.py`
+# actually reach a collector. Without this package the spans degrade
+# to OTel's NonRecordingSpan (near-zero cost).
+
+[project]
+name = "agentforge-otel"
+version = "0.2.3"
+description = "OpenTelemetry tracing + metrics emitter for AgentForge"
+readme = "README.md"
+requires-python = ">=3.13"
+license = "Apache-2.0"
+license-files = ["LICENSE"]
+authors = [
+    {name = "The AgentForge Authors"},
+]
+keywords = ["ai", "agent", "observability", "opentelemetry", "tracing"]
+classifiers = [
+    "Development Status :: 2 - Pre-Alpha",
+    "Intended Audience :: Developers",
+    "License :: OSI Approved :: Apache Software License",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.13",
+    "Topic :: System :: Monitoring",
+    "Typing :: Typed",
+]
+
+dependencies = [
+    "agentforge-core ~= 0.2.3",
+    "opentelemetry-sdk>=1.27",
+    "opentelemetry-exporter-otlp-proto-grpc>=1.27",
+]
+
+[project.urls]
+Homepage = "https://github.com/Scaffoldic/agentforge-py"
+Repository = "https://github.com/Scaffoldic/agentforge-py"
+Documentation = "https://github.com/Scaffoldic/agentforge-py"
+Changelog = "https://github.com/Scaffoldic/agentforge-py/blob/main/CHANGELOG.md"
+Issues = "https://github.com/Scaffoldic/agentforge-py/issues"
+
+# Entry-point registration — feat-010 will load by name from
+# `modules.observability` in agentforge.yaml.
+[project.entry-points."agentforge.hooks"]
+otel = "agentforge_otel.hook:OpenTelemetryHook"
+
+[build-system]
+requires = ["hatchling>=1.27"]
+build-backend = "hatchling.build"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/agentforge_otel"]
+
+[tool.hatch.build.targets.sdist]
+include = [
+    "src/agentforge_otel",
+    "tests",
+    "README.md",
+    "LICENSE",
+]

agentforge_otel-0.2.3/src/agentforge_otel/__init__.py

@@ -0,0 +1,13 @@
+"""OpenTelemetry tracing for AgentForge — feat-009.
+
+Public surface:
+- `OpenTelemetryHook` — configures the SDK provider + OTLP exporter
+  on construction, satisfies both `on_step` and `on_finish` hook
+  contracts via `__call__(step_or_result)` dispatch.
+"""
+
+from __future__ import annotations
+
+from agentforge_otel.hook import OpenTelemetryHook
+
+__all__ = ["OpenTelemetryHook"]

agentforge_otel-0.2.3/src/agentforge_otel/hook.py

@@ -0,0 +1,235 @@
+"""`OpenTelemetryHook` — SDK setup + per-step/per-finish hook impl.
+
+Construction installs the OTel SDK tracer provider with an OTLP gRPC
+exporter and a `TraceIdRatioBased` sampler. Idempotent — repeat
+construction reuses the existing provider.
+
+The hook itself satisfies both step and finish hook contracts via
+`__call__(payload)`:
+- `Step` -> annotate the current span (run span or whatever's active)
+  with per-step attributes.
+- `RunResult` -> the run span has already been closed by `Agent.run`'s
+  context manager; the hook just adds a final summary log via the
+  `agentforge.observability` logger for callers that want a compact
+  per-run line in their logs.
+
+For arg redaction: the hook scrubs values whose keys match
+`redact_fields` (case-insensitive substring match) before adding tool
+call args as span attributes.
+"""
+
+from __future__ import annotations
+
+import logging
+import re
+import threading
+from typing import Any
+
+from agentforge_core.values.state import RunResult, Step
+from opentelemetry import trace
+from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
+from opentelemetry.sdk.resources import SERVICE_NAME, Resource
+from opentelemetry.sdk.trace import TracerProvider
+from opentelemetry.sdk.trace.export import BatchSpanProcessor
+from opentelemetry.sdk.trace.sampling import TraceIdRatioBased
+
+_log = logging.getLogger("agentforge.observability")
+
+_DEFAULT_REDACT = ("api_key", "password", "secret", "token", "authorization")
+
+# Module-level lock so concurrent hook instantiations don't race on
+# the global tracer provider. The "installed" flag lives on a mutable
+# list to avoid a `global` declaration (PLW0603).
+_setup_lock = threading.Lock()
+_provider_installed: list[bool] = [False]
+
+
+class OpenTelemetryHook:
+    """Configures OTel SDK on construction; satisfies StepHook + FinishHook.
+
+    Args:
+        endpoint: OTLP gRPC endpoint (e.g. `http://otel-collector:4317`).
+            Defaults to the OTel SDK's own env-var defaults
+            (`OTEL_EXPORTER_OTLP_ENDPOINT`).
+        service_name: Resource attribute under `service.name`. Required.
+        sample_rate: `TraceIdRatioBased` rate in [0.0, 1.0]. Default 1.0
+            (every trace sampled).
+        redact_fields: Substring matches for sensitive field names.
+            Defaults to common API-key / password / token names; pass
+            your own to override.
+    """
+
+    def __init__(
+        self,
+        *,
+        endpoint: str | None = None,
+        service_name: str = "agentforge",
+        sample_rate: float = 1.0,
+        redact_fields: tuple[str, ...] | None = None,
+        redact_value_patterns: tuple[str, ...] | None = None,
+    ) -> None:
+        if not 0.0 <= sample_rate <= 1.0:
+            raise ValueError(f"sample_rate must be in [0, 1]; got {sample_rate}")
+        if not service_name:
+            raise ValueError("service_name is required")
+
+        self._service_name = service_name
+        self._endpoint = endpoint
+        self._sample_rate = sample_rate
+        self._redact_fields = tuple(
+            f.lower() for f in (redact_fields if redact_fields is not None else _DEFAULT_REDACT)
+        )
+        # feat-009 v0.3 polish: content-based PII redaction. Patterns
+        # are compiled once at construction so per-step redaction
+        # stays cheap. Default None — content-based redaction is
+        # opt-in.
+        self._redact_value_patterns: tuple[re.Pattern[str], ...] = (
+            tuple(re.compile(p) for p in redact_value_patterns)
+            if redact_value_patterns is not None
+            else ()
+        )
+        _ensure_provider(
+            service_name=service_name,
+            endpoint=endpoint,
+            sample_rate=sample_rate,
+        )
+
+    @property
+    def service_name(self) -> str:
+        return self._service_name
+
+    @property
+    def redact_fields(self) -> tuple[str, ...]:
+        return self._redact_fields
+
+    @property
+    def redact_value_patterns(self) -> tuple[re.Pattern[str], ...]:
+        return self._redact_value_patterns
+
+    def __call__(self, payload: Step | RunResult) -> None:
+        """Hook entry point — dispatches on payload type."""
+        if isinstance(payload, Step):
+            self._on_step(payload)
+        else:
+            self._on_finish(payload)
+
+    # ------------------------------------------------------------------
+    # Step handling — annotate the current span.
+    # ------------------------------------------------------------------
+
+    def _on_step(self, step: Step) -> None:
+        span = trace.get_current_span()
+        # Span is the no-op NonRecordingSpan when nothing is active —
+        # safe to call set_attribute on it (it just discards).
+        span.add_event(
+            "agent.step",
+            attributes={
+                "agentforge.step.iteration": step.iteration,
+                "agentforge.step.kind": step.kind,
+                "agentforge.step.cost_usd": step.cost_usd,
+                "agentforge.step.tokens_in": step.tokens_in,
+                "agentforge.step.tokens_out": step.tokens_out,
+                "agentforge.step.duration_ms": step.duration_ms,
+            },
+        )
+        if step.tool_call is not None:
+            span.add_event(
+                "agent.tool_call",
+                attributes={
+                    "agentforge.tool.name": step.tool_call.name,
+                    "agentforge.tool.args": self._redact(dict(step.tool_call.arguments)),
+                },
+            )
+
+    # ------------------------------------------------------------------
+    # Finish handling — the run span already closed; log a summary.
+    # ------------------------------------------------------------------
+
+    def _on_finish(self, result: RunResult) -> None:
+        _log.info(
+            "run %s finished: %s (cost=$%.4f, tokens_in=%d, tokens_out=%d, steps=%d, %dms)",
+            result.run_id,
+            result.finish_reason,
+            result.cost_usd,
+            result.tokens_in,
+            result.tokens_out,
+            len(result.steps),
+            result.duration_ms,
+        )
+
+    # ------------------------------------------------------------------
+    # Redaction helper.
+    # ------------------------------------------------------------------
+
+    def _redact(self, args: dict[str, Any]) -> str:
+        """Stringify a dict for span attribution, masking sensitive
+        keys *and* values matching configured regex patterns.
+
+        Two redaction passes:
+        1. Key-based: case-insensitive substring match on the field
+           name against `redact_fields` (e.g. ``api_key``,
+           ``password``). The whole value gets masked.
+        2. Content-based (feat-009 v0.3 polish): for string values
+           that survived the key check, run each pattern in
+           `redact_value_patterns` against the stringified value;
+           any match masks the whole value.
+
+        Span attributes accept primitives + lists thereof; we render
+        the dict to a compact ``k=v`` form rather than fight the
+        schema.
+        """
+        parts: list[str] = []
+        for key, value in args.items():
+            if any(token in key.lower() for token in self._redact_fields):
+                parts.append(f"{key}=<redacted>")
+                continue
+            if self._value_matches_pattern(value):
+                parts.append(f"{key}=<redacted>")
+                continue
+            parts.append(f"{key}={value!r}")
+        return ", ".join(parts)
+
+    def _value_matches_pattern(self, value: Any) -> bool:
+        """True if any configured regex pattern matches the value
+        text. Non-string values are stringified first so payloads
+        like ``42``, ``True``, or nested dicts are still scanned."""
+        if not self._redact_value_patterns:
+            return False
+        text = value if isinstance(value, str) else str(value)
+        return any(p.search(text) for p in self._redact_value_patterns)
+
+
+def _ensure_provider(
+    *,
+    service_name: str,
+    endpoint: str | None,
+    sample_rate: float,
+) -> None:
+    """Install the SDK provider + OTLP exporter once per process.
+
+    Idempotent: subsequent calls (with potentially-different config)
+    keep the first-installed provider. Re-installing would silently
+    drop spans already in-flight on the previous provider, so we
+    don't do that — first wins.
+    """
+    with _setup_lock:
+        if _provider_installed[0]:
+            return
+        # If the user has already installed a TracerProvider themselves
+        # (rare but possible — they wired OTel manually), respect it.
+        existing = trace.get_tracer_provider()
+        if isinstance(existing, TracerProvider):
+            _provider_installed[0] = True
+            return
+        resource = Resource.create({SERVICE_NAME: service_name})
+        provider = TracerProvider(
+            resource=resource,
+            sampler=TraceIdRatioBased(sample_rate),
+        )
+        exporter_kwargs: dict[str, Any] = {}
+        if endpoint is not None:
+            exporter_kwargs["endpoint"] = endpoint
+        exporter = OTLPSpanExporter(**exporter_kwargs)
+        provider.add_span_processor(BatchSpanProcessor(exporter))
+        trace.set_tracer_provider(provider)
+        _provider_installed[0] = True

agentforge_otel-0.2.3/tests/unit/test_hook.py

@@ -0,0 +1,478 @@
+"""Unit tests for `agentforge_otel.hook.OpenTelemetryHook`.
+
+Uses OTel's `InMemorySpanExporter` so we can assert on the actual
+spans the framework + hook produce, without sending traffic anywhere.
+"""
+
+from __future__ import annotations
+
+import logging
+
+import agentforge_otel.hook as hook_mod
+import pytest
+from agentforge import Agent
+from agentforge_core.contracts.strategy import ReasoningStrategy
+from agentforge_core.values.messages import ToolCall
+from agentforge_core.values.state import AgentState, RunResult, Step
+from agentforge_otel import OpenTelemetryHook
+from opentelemetry import trace
+from opentelemetry.sdk.resources import Resource
+from opentelemetry.sdk.trace import TracerProvider
+from opentelemetry.sdk.trace.export import SimpleSpanProcessor
+from opentelemetry.sdk.trace.export.in_memory_span_exporter import InMemorySpanExporter
+
+
+@pytest.fixture(autouse=True)
+def _reset_provider_state():
+    """Reset the module-level "provider installed" flag + tracer
+    provider so each test starts clean. OTel's `set_tracer_provider`
+    refuses to swap once set, so we reset the internals.
+    """
+    hook_mod._provider_installed[0] = False
+    trace._TRACER_PROVIDER_SET_ONCE = trace.Once()  # type: ignore[attr-defined]
+    trace._TRACER_PROVIDER = None  # type: ignore[attr-defined]
+
+
+def _install_inmemory_provider() -> InMemorySpanExporter:
+    """Replace OTel's provider with one that exports to an in-memory
+    list. Returns the exporter so tests can inspect emitted spans."""
+    exporter = InMemorySpanExporter()
+    provider = TracerProvider(resource=Resource.create({"service.name": "test"}))
+    provider.add_span_processor(SimpleSpanProcessor(exporter))
+    trace.set_tracer_provider(provider)
+    hook_mod._provider_installed[0] = True
+    return exporter
+
+
+# --- construction ---------------------------------------------------
+
+
+def test_invalid_sample_rate_rejected():
+    with pytest.raises(ValueError, match="sample_rate"):
+        OpenTelemetryHook(service_name="t", sample_rate=1.5)
+    with pytest.raises(ValueError, match="sample_rate"):
+        OpenTelemetryHook(service_name="t", sample_rate=-0.1)
+
+
+def test_missing_service_name_rejected():
+    with pytest.raises(ValueError, match="service_name"):
+        OpenTelemetryHook(service_name="")
+
+
+def test_construction_idempotent_does_not_replace_provider():
+    """If a provider is already installed, the hook doesn't clobber it."""
+    exporter = _install_inmemory_provider()
+    existing_provider = trace.get_tracer_provider()
+
+    OpenTelemetryHook(service_name="agent-1", endpoint="http://nowhere:4317")
+    # Provider unchanged — installation respected the existing one.
+    assert trace.get_tracer_provider() is existing_provider
+    assert exporter is not None  # exporter still wired
+
+
+# --- step events ---------------------------------------------------
+
+
+def test_step_event_added_to_current_span():
+    exporter = _install_inmemory_provider()
+    h = OpenTelemetryHook(service_name="test", endpoint="http://localhost:4317")
+
+    tracer = trace.get_tracer("agentforge")
+    with tracer.start_as_current_span("agent.run"):
+        h(
+            Step(
+                iteration=0,
+                kind="observe",
+                content="hello",
+                tokens_in=10,
+                tokens_out=20,
+                cost_usd=0.001,
+                duration_ms=42,
+            )
+        )
+
+    finished = exporter.get_finished_spans()
+    assert len(finished) == 1
+    events = list(finished[0].events)
+    step_events = [e for e in events if e.name == "agent.step"]
+    assert len(step_events) == 1
+    attrs = dict(step_events[0].attributes or {})
+    assert attrs["agentforge.step.iteration"] == 0
+    assert attrs["agentforge.step.kind"] == "observe"
+    assert attrs["agentforge.step.cost_usd"] == pytest.approx(0.001)
+    assert attrs["agentforge.step.tokens_in"] == 10
+
+
+def test_tool_call_event_added_with_redaction():
+    exporter = _install_inmemory_provider()
+    h = OpenTelemetryHook(service_name="test", endpoint="http://localhost:4317")
+
+    tracer = trace.get_tracer("agentforge")
+    with tracer.start_as_current_span("agent.run"):
+        h(
+            Step(
+                iteration=0,
+                kind="act",
+                content="call",
+                tool_call=ToolCall(
+                    id="01TEST",
+                    name="charge_customer",
+                    arguments={
+                        "customer_id": "cust_42",
+                        "api_key": "sk-secret-xyz",
+                        "amount_cents": 1000,
+                    },
+                ),
+            )
+        )
+
+    finished = exporter.get_finished_spans()
+    events = [e for e in finished[0].events if e.name == "agent.tool_call"]
+    assert len(events) == 1
+    attrs = dict(events[0].attributes or {})
+    args_str = str(attrs["agentforge.tool.args"])
+    assert "customer_id=" in args_str
+    assert "amount_cents=" in args_str
+    # api_key value redacted; key still visible.
+    assert "api_key=<redacted>" in args_str
+    assert "sk-secret-xyz" not in args_str
+
+
+def test_custom_redact_fields_take_effect():
+    exporter = _install_inmemory_provider()
+    h = OpenTelemetryHook(
+        service_name="test", endpoint="http://localhost:4317", redact_fields=("ssn",)
+    )
+
+    tracer = trace.get_tracer("agentforge")
+    with tracer.start_as_current_span("agent.run"):
+        h(
+            Step(
+                iteration=0,
+                kind="act",
+                content="x",
+                tool_call=ToolCall(
+                    id="01T",
+                    name="lookup",
+                    arguments={"ssn": "123-45-6789", "api_key": "still-visible"},
+                ),
+            )
+        )
+
+    finished = exporter.get_finished_spans()
+    event = next(e for e in finished[0].events if e.name == "agent.tool_call")
+    args_str = str(dict(event.attributes or {})["agentforge.tool.args"])
+    assert "ssn=<redacted>" in args_str
+    # api_key no longer in the redact list — visible.
+    assert "still-visible" in args_str
+
+
+# --- finish handling ----------------------------------------------
+
+
+def test_finish_emits_summary_log(caplog):
+    caplog.set_level(logging.INFO, logger="agentforge.observability")
+    h = OpenTelemetryHook(service_name="test", endpoint="http://localhost:4317")
+
+    result = RunResult(
+        output="done",
+        cost_usd=0.05,
+        tokens_in=100,
+        tokens_out=50,
+        run_id="01HXFINISH",
+        duration_ms=1234,
+        finish_reason="completed",
+    )
+    h(result)
+
+    msgs = [r.message for r in caplog.records]
+    assert any("01HXFINISH" in m and "completed" in m and "$0.0500" in m for m in msgs)
+
+
+# --- public properties --------------------------------------------
+
+
+def test_public_properties():
+    h = OpenTelemetryHook(
+        service_name="my-agent",
+        endpoint="http://localhost:4317",
+        redact_fields=("custom",),
+    )
+    assert h.service_name == "my-agent"
+    assert h.redact_fields == ("custom",)
+
+
+# --- end-to-end with Agent --------------------------------------
+
+
+@pytest.mark.asyncio
+async def test_end_to_end_root_span_from_agent_run():
+    """Construct an Agent, run it with the OTel hook installed, assert
+    the framework's `agent.run` root span landed in the exporter with
+    the right attributes."""
+    exporter = _install_inmemory_provider()
+
+    class _OneStep(ReasoningStrategy):
+        async def run(self, state: AgentState) -> AgentState:
+            state.steps.append(Step(iteration=0, kind="observe", content="hi"))
+            return state
+
+    otel = OpenTelemetryHook(service_name="test", endpoint="http://localhost:4317")
+    async with Agent(strategy=_OneStep(), on_step=otel, on_finish=otel) as agent:
+        await agent.run("the task")
+
+    finished = exporter.get_finished_spans()
+    run_spans = [s for s in finished if s.name == "agent.run"]
+    assert len(run_spans) == 1
+    attrs = dict(run_spans[0].attributes or {})
+    assert attrs["agentforge.task"] == "the task"
+    assert attrs["agentforge.finish_reason"] == "completed"
+    assert attrs["agentforge.n_steps"] == 1
+    # The step hook annotated the span with an event.
+    step_events = [e for e in run_spans[0].events if e.name == "agent.step"]
+    assert len(step_events) == 1
+
+
+@pytest.mark.asyncio
+async def test_child_span_tree_via_react_loop():
+    """Drive an Agent with the built-in ReActLoop strategy and a
+    scripted FakeLLM that triggers one tool_call. Assert the full
+    OTel span tree lands:
+      agent.run
+        ├── strategy.iteration (iter 0)
+        │   ├── llm.call
+        │   └── tool.<name>
+        └── strategy.iteration (iter 1)
+            └── llm.call
+    """
+    from agentforge._testing import FakeLLMClient  # noqa: PLC0415
+    from agentforge.strategies.react import ReActLoop  # noqa: PLC0415
+    from agentforge_core.contracts.tool import Tool  # noqa: PLC0415
+    from agentforge_core.values.messages import LLMResponse, TokenUsage  # noqa: PLC0415
+    from pydantic import BaseModel  # noqa: PLC0415
+
+    exporter = _install_inmemory_provider()
+
+    class _SearchArgs(BaseModel):
+        query: str
+
+    class _SearchTool(Tool):
+        name = "search"
+        description = "echoes the query"
+        input_schema = _SearchArgs
+
+        async def run(self, query: str) -> str:
+            return f"got {query!r}"
+
+    fake = FakeLLMClient(
+        responses=[
+            LLMResponse(
+                content="I'll search.",
+                stop_reason="tool_use",
+                tool_calls=(ToolCall(id="t-1", name="search", arguments={"query": "x"}),),
+                usage=TokenUsage(input_tokens=5, output_tokens=3),
+                cost_usd=0.0,
+                model="fake",
+                provider="fake",
+            ),
+            LLMResponse(
+                content="done",
+                stop_reason="end_turn",
+                usage=TokenUsage(input_tokens=4, output_tokens=2),
+                cost_usd=0.0,
+                model="fake",
+                provider="fake",
+            ),
+        ],
+    )
+    otel = OpenTelemetryHook(service_name="test", endpoint="http://localhost:4317")
+    async with Agent(
+        model=fake,
+        tools=[_SearchTool()],
+        strategy=ReActLoop(),
+        on_step=otel,
+        on_finish=otel,
+    ) as agent:
+        await agent.run("find x")
+
+    finished = exporter.get_finished_spans()
+    by_name: dict[str, list] = {}
+    for s in finished:
+        by_name.setdefault(s.name, []).append(s)
+
+    # Root span
+    assert len(by_name["agent.run"]) == 1
+    # Two iterations: one with the tool_call, one terminating
+    assert len(by_name["strategy.iteration"]) == 2
+    # Two LLM calls (one per iteration)
+    assert len(by_name["llm.call"]) == 2
+    # One tool execution
+    assert len(by_name["tool.search"]) == 1
+    # llm.call spans carry provider + token attributes
+    llm_attrs = dict(by_name["llm.call"][0].attributes or {})
+    assert llm_attrs["agentforge.llm.provider"] == "fake"
+    assert llm_attrs["agentforge.llm.tokens_in"] == 5
+
+
+@pytest.mark.asyncio
+async def test_child_span_tree_via_tot_strategy():
+    """Drive an Agent with the built-in ``TreeOfThoughts`` strategy
+    and assert ``strategy.iteration`` spans land under ``agent.run``
+    with ``agentforge.strategy=tot``.
+    """
+    from agentforge._testing import FakeLLMClient  # noqa: PLC0415
+    from agentforge.strategies.tot import TreeOfThoughts  # noqa: PLC0415
+    from agentforge_core.values.messages import LLMResponse, TokenUsage  # noqa: PLC0415
+
+    exporter = _install_inmemory_provider()
+
+    def _resp(content: str) -> LLMResponse:
+        return LLMResponse(
+            content=content,
+            stop_reason="end_turn",
+            usage=TokenUsage(input_tokens=5, output_tokens=3),
+            cost_usd=0.0,
+            model="fake",
+            provider="fake",
+        )
+
+    fake = FakeLLMClient(
+        responses=[
+            _resp('{"thoughts": [{"id": "t1", "content": "thought 1"}]}'),
+            _resp('{"scores": [{"branch_id": "t1", "score": 0.9, "reasoning": "ok"}]}'),
+            _resp("Final answer."),
+        ]
+    )
+    otel = OpenTelemetryHook(service_name="test", endpoint="http://localhost:4317")
+    async with Agent(
+        model=fake,
+        tools=[],
+        strategy=TreeOfThoughts(branch_factor=1, depth=1, score_threshold=0.5),
+        on_step=otel,
+        on_finish=otel,
+    ) as agent:
+        await agent.run("solve x")
+
+    finished = exporter.get_finished_spans()
+    by_name: dict[str, list] = {}
+    for s in finished:
+        by_name.setdefault(s.name, []).append(s)
+
+    assert len(by_name["agent.run"]) == 1
+    assert len(by_name["strategy.iteration"]) == 1
+    attrs = dict(by_name["strategy.iteration"][0].attributes or {})
+    assert attrs["agentforge.strategy"] == "tot"
+    assert attrs["agentforge.iteration"] == 0
+
+
+@pytest.mark.asyncio
+async def test_child_span_tree_via_multi_agent_strategy():
+    """Drive an Agent with the built-in ``MultiAgentSupervisor`` and
+    assert ``strategy.iteration`` spans land with
+    ``agentforge.strategy=multi_agent``.
+    """
+    from agentforge._testing import FakeLLMClient  # noqa: PLC0415
+    from agentforge.strategies._base import get_runtime  # noqa: PLC0415
+    from agentforge.strategies.multi_agent import MultiAgentSupervisor  # noqa: PLC0415
+    from agentforge_core.contracts.strategy import ReasoningStrategy  # noqa: PLC0415
+    from agentforge_core.values.messages import LLMResponse, TokenUsage  # noqa: PLC0415
+    from agentforge_core.values.state import Step as StepCls  # noqa: PLC0415
+
+    exporter = _install_inmemory_provider()
+
+    class _Worker(ReasoningStrategy):
+        async def run(self, state):
+            get_runtime(state).budget.check()
+            state.steps.append(StepCls(iteration=1, kind="synthesize", content="ok", cost_usd=0.0))
+            return state
+
+    def _resp(content: str) -> LLMResponse:
+        return LLMResponse(
+            content=content,
+            stop_reason="end_turn",
+            usage=TokenUsage(input_tokens=5, output_tokens=3),
+            cost_usd=0.0,
+            model="fake",
+            provider="fake",
+        )
+
+    fake = FakeLLMClient(
+        responses=[
+            _resp('{"assignments": [{"worker": "a", "task": "subtask 1"}]}'),
+            _resp("aggregated answer."),
+        ]
+    )
+    otel = OpenTelemetryHook(service_name="test", endpoint="http://localhost:4317")
+    async with Agent(
+        model=fake,
+        tools=[],
+        strategy=MultiAgentSupervisor(workers={"a": _Worker()}, max_rounds=1),
+        on_step=otel,
+        on_finish=otel,
+    ) as agent:
+        await agent.run("big task")
+
+    finished = exporter.get_finished_spans()
+    by_name: dict[str, list] = {}
+    for s in finished:
+        by_name.setdefault(s.name, []).append(s)
+
+    assert len(by_name["agent.run"]) == 1
+    assert len(by_name["strategy.iteration"]) == 1
+    attrs = dict(by_name["strategy.iteration"][0].attributes or {})
+    assert attrs["agentforge.strategy"] == "multi_agent"
+    assert attrs["agentforge.iteration"] == 0
+
+
+# --- content-based PII redaction (feat-009 v0.3 polish) --------
+
+
+def test_redact_value_patterns_masks_matching_values() -> None:
+    """Values matching any regex in `redact_value_patterns` get
+    replaced wholesale — even when the key isn't in
+    `redact_fields`."""
+    h = OpenTelemetryHook(
+        service_name="test",
+        endpoint="http://localhost:4317",
+        redact_value_patterns=(
+            r"\b\d{3}-\d{2}-\d{4}\b",  # US SSN
+            r"\b\d{16}\b",  # Credit-card-like 16-digit run
+            r"[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}",  # email
+        ),
+    )
+    rendered = h._redact(
+        {
+            "ssn": "123-45-6789",
+            "card": "4111111111111111",
+            "email": "alice@example.com",
+            "innocent": "hello world",
+            "count": 7,
+        }
+    )
+    assert "ssn=<redacted>" in rendered
+    assert "card=<redacted>" in rendered
+    assert "email=<redacted>" in rendered
+    # Plain integers + harmless strings pass through.
+    assert "innocent='hello world'" in rendered
+    assert "count=7" in rendered
+
+
+def test_redact_value_patterns_default_none_is_passthrough() -> None:
+    """Without `redact_value_patterns`, behaviour is identical to
+    v0.1 (key-only redaction)."""
+    h = OpenTelemetryHook(service_name="test", endpoint="http://localhost:4317")
+    assert h.redact_value_patterns == ()
+    rendered = h._redact({"email": "alice@example.com", "api_key": "shh"})
+    assert "email='alice@example.com'" in rendered
+    assert "api_key=<redacted>" in rendered
+
+
+def test_redact_key_match_takes_precedence_over_value_match() -> None:
+    h = OpenTelemetryHook(
+        service_name="test",
+        endpoint="http://localhost:4317",
+        redact_value_patterns=(r"unused-pattern",),
+    )
+    rendered = h._redact({"api_key": "abc"})
+    # Key matches first; we never even check the value patterns.
+    assert "api_key=<redacted>" in rendered