agent-panorama 0.1.0__tar.gz

agent_panorama-0.1.0/.github/workflows/ci.yml

@@ -0,0 +1,40 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+permissions:
+  contents: read
+
+jobs:
+  test:
+    name: Lint & test (py${{ matrix.python-version }})
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.11", "3.12"]
+
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v5
+        with:
+          enable-cache: true
+
+      - name: Sync dependencies
+        run: uv sync --extra dev --python ${{ matrix.python-version }}
+
+      - name: Ruff lint
+        run: uv run ruff check src tests
+
+      - name: Ruff format check
+        run: uv run ruff format --check src tests
+
+      - name: Run tests
+        run: uv run python tests/run_all_tests.py

agent_panorama-0.1.0/.github/workflows/publish.yml

@@ -0,0 +1,29 @@
+name: Publish to PyPI
+
+on:
+  release:
+    types: [published]
+
+permissions:
+  contents: read
+
+jobs:
+  publish:
+    name: Build and publish to PyPI
+    runs-on: ubuntu-latest
+    environment: pypi
+    permissions:
+      id-token: write  # required for PyPI Trusted Publishing (OIDC)
+
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v5
+
+      - name: Build sdist and wheel
+        run: uv build
+
+      - name: Publish to PyPI (trusted publishing)
+        run: uv publish

agent_panorama-0.1.0/.gitignore

@@ -0,0 +1,32 @@
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+.eggs/
+build/
+dist/
+.venv/
+venv/
+
+# Tooling
+.mypy_cache/
+.ruff_cache/
+.pytest_cache/
+
+# Generated reports
+report.md
+report.html
+/report/
+
+# Real trace data dropped in for local validation (keep the instructions file)
+/traces/*
+!/traces/PUT_YOUR_TRACE_HERE.md
+
+# Never commit raw trace exports anywhere in the tree (may contain private data)
+trace-*.json
+*trace_export*.json
+
+# OS / editors
+.DS_Store
+.idea/
+.vscode/

agent_panorama-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 agent-panorama contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

agent_panorama-0.1.0/PKG-INFO

@@ -0,0 +1,211 @@
+Metadata-Version: 2.4
+Name: agent-panorama
+Version: 0.1.0
+Summary: Turn Langfuse/LangSmith agent traces into human-readable Agent Activity Reports (Markdown + HTML).
+Project-URL: Homepage, https://github.com/Idank96/agent-panorama
+Project-URL: Repository, https://github.com/Idank96/agent-panorama
+Project-URL: Issues, https://github.com/Idank96/agent-panorama/issues
+Author: agent-panorama contributors
+License: MIT
+License-File: LICENSE
+Keywords: agents,langfuse,langsmith,llm,observability,reporting
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Software Development :: Libraries
+Requires-Python: >=3.10
+Requires-Dist: click>=8.1
+Requires-Dist: jinja2>=3.1
+Requires-Dist: pyyaml>=6.0
+Provides-Extra: dev
+Requires-Dist: mypy>=1.10; extra == 'dev'
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: ruff>=0.6; extra == 'dev'
+Requires-Dist: types-pyyaml>=6.0; extra == 'dev'
+Description-Content-Type: text/markdown
+
+<p align="center">
+  <img src="assets/logo.png" alt="agent-panorama" width="320">
+</p>
+
+<h1 align="center">agent-panorama</h1>
+
+<p align="center">
+  <a href="https://github.com/Idank96/agent-panorama/actions/workflows/ci.yml"><img src="https://github.com/Idank96/agent-panorama/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
+</p>
+
+Turn raw LLM agent traces into a **human-readable Agent Activity Report** that a
+non-engineer can actually read. Point it at a Langfuse (or LangSmith) trace
+export and get clean Markdown + a self-contained HTML report that explains, in
+business language, what your agents did, what they decided, and anything that
+looks off.
+
+## Why
+
+Traces are great for engineers and terrible for everyone else. `agent-panorama`
+translates tool calls, retries, token usage, and errors into plain English. It
+also pulls the real user request and final answer out of LangGraph/LangChain
+`messages` payloads, so the report reads like a story, not a JSON dump:
+
+- `get_weather({"city": "Paris"})` → **"Looked up the weather"**
+- 3 failed model calls → **"High retry count: 3 failed attempts before completing."**
+- `human_handoff(...)` → run outcome **human-escalated**
+
+> Cost/USD estimation is intentionally out of scope for now — the report reports
+> token usage, not dollars.
+
+## Install
+
+```bash
+pip install agent-panorama
+# or, for local development:
+uv pip install -e ".[dev]"
+```
+
+Requires Python 3.10+. Dependencies are intentionally minimal: `click`,
+`jinja2`, `pyyaml`.
+
+## CLI usage
+
+```bash
+agent-panorama generate --input traces.json --output ./report --format html
+```
+
+Options:
+
+| Option | Description |
+| --- | --- |
+| `--input` | Path to the Langfuse/LangSmith JSON export (required). |
+| `--output` | Output directory (default `./report`). |
+| `--format` | `md`, `html`, or `both` (default `both`). |
+| `--input-type` | `langfuse` or `langsmith` (default `langfuse`). |
+| `--config` | Optional YAML config (tool naming, thresholds). |
+
+Try it on the bundled example:
+
+```bash
+agent-panorama generate --input examples/langfuse_traces.json --output ./report
+```
+
+## Library usage
+
+```python
+from agent_panorama import generate_report
+
+report = generate_report(
+    "traces.json",
+    output_dir="./report",
+    formats=["md", "html"],
+    input_type="langfuse",
+    config="config.yaml",  # optional
+)
+
+print(report.total_runs, report.total_tokens)
+```
+
+`generate_report` returns the in-memory `Report`, so you can also inspect runs,
+the decision log, and anomalies programmatically without touching disk (use
+`build_report_from_file` if you want the report without writing files).
+
+## What's in a report
+
+- **Summary** — time range, total runs, total actions, total tokens.
+- **Per-agent section** — what it was asked to do, what it decided/did (tool calls
+  in plain English), final outcome, and a confidence signal (retries / fallback).
+- **Decision log** — a sortable table of every consequential action: timestamp,
+  agent, action, parameters summarized in plain English, outcome.
+- **Anomalies** — high retry counts, slow runs, high activity, errors, fallbacks.
+
+## Configuration
+
+All configuration is optional. See [`config.example.yaml`](config.example.yaml)
+for the full set. Highlights:
+
+```yaml
+tool_descriptions:
+  get_weather: "Looked up the weather"
+
+consequential_tools: [send_email, human_handoff]
+escalation_tools: [human_handoff, handoff_to_agent]
+
+anomaly_thresholds:
+  max_retries: 2
+  max_latency_seconds: 30
+  max_tool_calls: 15
+```
+
+## Supported inputs
+
+- **Langfuse** trace exports — a single trace dict, the single-trace
+  `{"trace": {...}, "observations": [...]}` shape, a list of traces, or the
+  `{"data": [...]}` list-API shape. Tool calls are read from `TOOL`
+  observations (falling back to tool spans), and from `toolCalls` / OpenAI-style
+  `tool_calls` declared on generations.
+- **LangSmith** run exports — a flat list (or `{"runs": [...]}`) of run nodes;
+  each root run is flattened into one agent run.
+
+Token usage is read from the trace (`inputUsage`/`outputUsage` or
+`usage`/`usage_metadata`). Dollar-cost estimation is intentionally out of scope.
+
+## Roadmap
+
+`agent-panorama` starts as a report generator and is growing into an **oversight
+layer for fleets of agents** — a single pane of glass for everything your agents
+did, decided, and got wrong. More than logs, across more than one agent.
+
+**✅ v0.1 — Read one run clearly _(today)_**
+- Langfuse + LangSmith trace ingestion
+- Plain-language per-agent summaries, decision log, anomalies
+- Markdown + self-contained HTML output; CLI and library API
+
+**🔜 v0.2 — See the whole fleet (the panorama view)**
+- A unified **cross-agent activity feed** — one scannable timeline of what every
+  agent did, in plain English:
+
+  ```text
+  Agent Activity — May 28, 14:30–15:00
+
+  research-assistant    → searched the web, summarized 3 papers            ✓ success
+  scheduling-assistant  → checked the calendar, handed the task to a human ⤴ escalated
+  weather-assistant     → looked up the weather (retried once), emailed it ✓ success
+  billing-agent         → issued 2 refunds, flagged 1 for review           ⚠ anomaly
+  ```
+- Aggregate many traces into one report (by session, time window, or file glob)
+- Per-agent rollups: runs, actions, success / escalation / retry rates
+- Cross-agent decision log spanning every agent in the window
+
+**📈 v0.3 — Trends & regressions**
+- Track rates over time, not just a point-in-time snapshot
+- Flag regressions (escalations or retries spiking vs. a baseline)
+- Period-over-period comparison ("this week vs. last")
+
+**🔌 v0.4 — More sources & deeper detail**
+- OpenTelemetry / OpenInference and raw OpenAI-style logs
+- Optionally fetch full input/output from the Langfuse API to enrich
+  decision-log parameters
+- Pluggable parser interface for custom trace formats
+
+**🎯 The vision — Continuous oversight**
+- A live dashboard: the activity feed above, always-on, filterable by agent /
+  outcome / time
+- Scheduled/continuous reports instead of one-off runs
+- Accountability views a non-engineer can sign off on (what happened, what needs
+  a human)
+- Alerting on anomalies across the fleet
+
+> Have a use case or a trace format you want supported? Open an issue.
+
+## Development
+
+```bash
+uv pip install -e ".[dev]"
+python tests/run_all_tests.py     # run the full suite
+ruff check . && ruff format --check .
+```
+
+## License
+
+MIT — see [LICENSE](LICENSE).

agent_panorama-0.1.0/README.md

@@ -0,0 +1,182 @@
+<p align="center">
+  <img src="assets/logo.png" alt="agent-panorama" width="320">
+</p>
+
+<h1 align="center">agent-panorama</h1>
+
+<p align="center">
+  <a href="https://github.com/Idank96/agent-panorama/actions/workflows/ci.yml"><img src="https://github.com/Idank96/agent-panorama/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
+</p>
+
+Turn raw LLM agent traces into a **human-readable Agent Activity Report** that a
+non-engineer can actually read. Point it at a Langfuse (or LangSmith) trace
+export and get clean Markdown + a self-contained HTML report that explains, in
+business language, what your agents did, what they decided, and anything that
+looks off.
+
+## Why
+
+Traces are great for engineers and terrible for everyone else. `agent-panorama`
+translates tool calls, retries, token usage, and errors into plain English. It
+also pulls the real user request and final answer out of LangGraph/LangChain
+`messages` payloads, so the report reads like a story, not a JSON dump:
+
+- `get_weather({"city": "Paris"})` → **"Looked up the weather"**
+- 3 failed model calls → **"High retry count: 3 failed attempts before completing."**
+- `human_handoff(...)` → run outcome **human-escalated**
+
+> Cost/USD estimation is intentionally out of scope for now — the report reports
+> token usage, not dollars.
+
+## Install
+
+```bash
+pip install agent-panorama
+# or, for local development:
+uv pip install -e ".[dev]"
+```
+
+Requires Python 3.10+. Dependencies are intentionally minimal: `click`,
+`jinja2`, `pyyaml`.
+
+## CLI usage
+
+```bash
+agent-panorama generate --input traces.json --output ./report --format html
+```
+
+Options:
+
+| Option | Description |
+| --- | --- |
+| `--input` | Path to the Langfuse/LangSmith JSON export (required). |
+| `--output` | Output directory (default `./report`). |
+| `--format` | `md`, `html`, or `both` (default `both`). |
+| `--input-type` | `langfuse` or `langsmith` (default `langfuse`). |
+| `--config` | Optional YAML config (tool naming, thresholds). |
+
+Try it on the bundled example:
+
+```bash
+agent-panorama generate --input examples/langfuse_traces.json --output ./report
+```
+
+## Library usage
+
+```python
+from agent_panorama import generate_report
+
+report = generate_report(
+    "traces.json",
+    output_dir="./report",
+    formats=["md", "html"],
+    input_type="langfuse",
+    config="config.yaml",  # optional
+)
+
+print(report.total_runs, report.total_tokens)
+```
+
+`generate_report` returns the in-memory `Report`, so you can also inspect runs,
+the decision log, and anomalies programmatically without touching disk (use
+`build_report_from_file` if you want the report without writing files).
+
+## What's in a report
+
+- **Summary** — time range, total runs, total actions, total tokens.
+- **Per-agent section** — what it was asked to do, what it decided/did (tool calls
+  in plain English), final outcome, and a confidence signal (retries / fallback).
+- **Decision log** — a sortable table of every consequential action: timestamp,
+  agent, action, parameters summarized in plain English, outcome.
+- **Anomalies** — high retry counts, slow runs, high activity, errors, fallbacks.
+
+## Configuration
+
+All configuration is optional. See [`config.example.yaml`](config.example.yaml)
+for the full set. Highlights:
+
+```yaml
+tool_descriptions:
+  get_weather: "Looked up the weather"
+
+consequential_tools: [send_email, human_handoff]
+escalation_tools: [human_handoff, handoff_to_agent]
+
+anomaly_thresholds:
+  max_retries: 2
+  max_latency_seconds: 30
+  max_tool_calls: 15
+```
+
+## Supported inputs
+
+- **Langfuse** trace exports — a single trace dict, the single-trace
+  `{"trace": {...}, "observations": [...]}` shape, a list of traces, or the
+  `{"data": [...]}` list-API shape. Tool calls are read from `TOOL`
+  observations (falling back to tool spans), and from `toolCalls` / OpenAI-style
+  `tool_calls` declared on generations.
+- **LangSmith** run exports — a flat list (or `{"runs": [...]}`) of run nodes;
+  each root run is flattened into one agent run.
+
+Token usage is read from the trace (`inputUsage`/`outputUsage` or
+`usage`/`usage_metadata`). Dollar-cost estimation is intentionally out of scope.
+
+## Roadmap
+
+`agent-panorama` starts as a report generator and is growing into an **oversight
+layer for fleets of agents** — a single pane of glass for everything your agents
+did, decided, and got wrong. More than logs, across more than one agent.
+
+**✅ v0.1 — Read one run clearly _(today)_**
+- Langfuse + LangSmith trace ingestion
+- Plain-language per-agent summaries, decision log, anomalies
+- Markdown + self-contained HTML output; CLI and library API
+
+**🔜 v0.2 — See the whole fleet (the panorama view)**
+- A unified **cross-agent activity feed** — one scannable timeline of what every
+  agent did, in plain English:
+
+  ```text
+  Agent Activity — May 28, 14:30–15:00
+
+  research-assistant    → searched the web, summarized 3 papers            ✓ success
+  scheduling-assistant  → checked the calendar, handed the task to a human ⤴ escalated
+  weather-assistant     → looked up the weather (retried once), emailed it ✓ success
+  billing-agent         → issued 2 refunds, flagged 1 for review           ⚠ anomaly
+  ```
+- Aggregate many traces into one report (by session, time window, or file glob)
+- Per-agent rollups: runs, actions, success / escalation / retry rates
+- Cross-agent decision log spanning every agent in the window
+
+**📈 v0.3 — Trends & regressions**
+- Track rates over time, not just a point-in-time snapshot
+- Flag regressions (escalations or retries spiking vs. a baseline)
+- Period-over-period comparison ("this week vs. last")
+
+**🔌 v0.4 — More sources & deeper detail**
+- OpenTelemetry / OpenInference and raw OpenAI-style logs
+- Optionally fetch full input/output from the Langfuse API to enrich
+  decision-log parameters
+- Pluggable parser interface for custom trace formats
+
+**🎯 The vision — Continuous oversight**
+- A live dashboard: the activity feed above, always-on, filterable by agent /
+  outcome / time
+- Scheduled/continuous reports instead of one-off runs
+- Accountability views a non-engineer can sign off on (what happened, what needs
+  a human)
+- Alerting on anomalies across the fleet
+
+> Have a use case or a trace format you want supported? Open an issue.
+
+## Development
+
+```bash
+uv pip install -e ".[dev]"
+python tests/run_all_tests.py     # run the full suite
+ruff check . && ruff format --check .
+```
+
+## License
+
+MIT — see [LICENSE](LICENSE).

agent_panorama-0.1.0/config.example.yaml

@@ -0,0 +1,27 @@
+# agent-panorama configuration (all sections optional).
+#
+# Pass with:  agent-panorama generate --input traces.json --output ./report --config config.yaml
+
+# Map raw tool names to readable descriptions used in the report.
+tool_descriptions:
+  web_search: "Searched the web"
+  get_weather: "Looked up the weather"
+  send_email: "Sent an email"
+  human_handoff: "Handed the task off to a person"
+
+# Tool names whose calls are considered "consequential" (side effects) and
+# therefore listed in the Decision Log. If omitted, every tool call is listed.
+consequential_tools:
+  - send_email
+  - human_handoff
+
+# Tool names that signal a human escalation (sets run outcome to "escalated").
+escalation_tools:
+  - human_handoff
+  - handoff_to_agent
+
+# Thresholds for the Anomalies section.
+anomaly_thresholds:
+  max_retries: 2          # more than this many retries is flagged
+  max_latency_seconds: 30 # runs slower than this are flagged
+  max_tool_calls: 15      # runs with more tool calls than this are flagged

agent_panorama-0.1.0/examples/langfuse_traces.json

@@ -0,0 +1,143 @@
+[
+  {
+    "id": "trace-001",
+    "name": "research-assistant",
+    "timestamp": "2026-05-20T09:15:00.000Z",
+    "input": {"question": "What are the top AI papers this week?"},
+    "output": {"content": "Here are three notable AI papers from this week, each with a short summary."},
+    "observations": [
+      {
+        "id": "obs-001-a",
+        "type": "GENERATION",
+        "name": "plan-response",
+        "startTime": "2026-05-20T09:15:01.000Z",
+        "endTime": "2026-05-20T09:15:03.500Z",
+        "model": "gpt-4o",
+        "usage": {"input": 800, "output": 120, "unit": "TOKENS"},
+        "level": "DEFAULT",
+        "output": {"tool_calls": [{"function": {"name": "web_search", "arguments": "{\"query\": \"top AI papers this week\"}"}}]}
+      },
+      {
+        "id": "obs-001-b",
+        "type": "SPAN",
+        "name": "web_search",
+        "startTime": "2026-05-20T09:15:03.600Z",
+        "endTime": "2026-05-20T09:15:04.200Z",
+        "input": {"query": "top AI papers this week"},
+        "output": {"results": 3},
+        "level": "DEFAULT"
+      },
+      {
+        "id": "obs-001-c",
+        "type": "SPAN",
+        "name": "summarize_text",
+        "startTime": "2026-05-20T09:15:04.300Z",
+        "endTime": "2026-05-20T09:15:04.900Z",
+        "input": {"documents": 3, "max_words": 80},
+        "output": {"summary": "three short summaries"},
+        "level": "DEFAULT"
+      },
+      {
+        "id": "obs-001-d",
+        "type": "GENERATION",
+        "name": "final-answer",
+        "startTime": "2026-05-20T09:15:05.000Z",
+        "endTime": "2026-05-20T09:15:06.400Z",
+        "model": "gpt-4o",
+        "usage": {"input": 1000, "output": 100, "unit": "TOKENS"},
+        "level": "DEFAULT",
+        "output": {"content": "Here are three notable AI papers from this week."}
+      }
+    ]
+  },
+  {
+    "id": "trace-002",
+    "name": "scheduling-assistant",
+    "timestamp": "2026-05-20T10:02:00.000Z",
+    "input": {"question": "Book a meeting with the design team next week."},
+    "output": {"content": "This request has been handed off to a human scheduler."},
+    "observations": [
+      {
+        "id": "obs-002-a",
+        "type": "GENERATION",
+        "name": "assess-request",
+        "startTime": "2026-05-20T10:02:01.000Z",
+        "endTime": "2026-05-20T10:02:04.000Z",
+        "model": "claude-3-5-sonnet-20241022",
+        "usage": {"input": 1200, "output": 200, "unit": "TOKENS"},
+        "level": "DEFAULT",
+        "output": {"content": "This needs a human to confirm availability."}
+      },
+      {
+        "id": "obs-002-b",
+        "type": "SPAN",
+        "name": "check_calendar",
+        "startTime": "2026-05-20T10:02:04.100Z",
+        "endTime": "2026-05-20T10:02:04.400Z",
+        "input": {"team": "design", "window": "next_week"},
+        "output": {"free_slots": 2},
+        "level": "DEFAULT"
+      },
+      {
+        "id": "obs-002-c",
+        "type": "SPAN",
+        "name": "human_handoff",
+        "startTime": "2026-05-20T10:02:04.500Z",
+        "endTime": "2026-05-20T10:02:04.800Z",
+        "input": {"reason": "needs human confirmation", "queue": "scheduling"},
+        "output": {"ticket_id": "hx-3320"},
+        "level": "DEFAULT"
+      }
+    ]
+  },
+  {
+    "id": "trace-003",
+    "name": "weather-assistant",
+    "timestamp": "2026-05-20T11:30:00.000Z",
+    "input": {"task": "Email me tomorrow's forecast for Paris."},
+    "output": {"content": "Forecast sent after recovering from a provider timeout."},
+    "observations": [
+      {
+        "id": "obs-003-a",
+        "type": "GENERATION",
+        "name": "plan",
+        "startTime": "2026-05-20T11:30:01.000Z",
+        "endTime": "2026-05-20T11:30:02.000Z",
+        "model": "gpt-4o-mini",
+        "usage": {"input": 500, "output": 70, "unit": "TOKENS"},
+        "level": "DEFAULT"
+      },
+      {
+        "id": "obs-003-b",
+        "type": "SPAN",
+        "name": "get_weather",
+        "startTime": "2026-05-20T11:30:02.100Z",
+        "endTime": "2026-05-20T11:30:14.100Z",
+        "input": {"city": "Paris", "provider": "primary"},
+        "output": {"error": "provider timeout"},
+        "level": "ERROR",
+        "statusMessage": "Weather provider timed out after 12s"
+      },
+      {
+        "id": "obs-003-c",
+        "type": "SPAN",
+        "name": "get_weather",
+        "startTime": "2026-05-20T11:30:14.200Z",
+        "endTime": "2026-05-20T11:30:15.400Z",
+        "input": {"city": "Paris", "provider": "backup"},
+        "output": {"temp_c": 18, "summary": "partly cloudy"},
+        "level": "DEFAULT"
+      },
+      {
+        "id": "obs-003-d",
+        "type": "SPAN",
+        "name": "send_email",
+        "startTime": "2026-05-20T11:30:15.500Z",
+        "endTime": "2026-05-20T11:30:16.000Z",
+        "input": {"to": "user", "subject": "Paris forecast for tomorrow"},
+        "output": {"sent": true},
+        "level": "DEFAULT"
+      }
+    ]
+  }
+]