Skip to main content

Observability tooling for agent harness sessions, imports, and reports.

Project description

Harness Observability Layer

Harness Observability Layer, or HOL, is a local-first dashboard for inspecting archived agent sessions from the current project.

Today HOL centers on a localhost experience:

  • it discovers Codex and Claude Code session archives that belong to the current repository
  • it normalizes those sessions into a shared event model in memory
  • it computes aggregate metrics, prompt-group rankings, turn rankings, and prescriptive insights
  • it serves a project dashboard locally at http://localhost

The main product experience is one command:

hol init

What HOL Does Today

HOL is built for questions like:

  • where did spend accumulate in this project
  • which prompts or turns were disproportionately expensive
  • whether long context or missing /clear is driving cost
  • whether model choice looks justified
  • which sessions deserve review first

The dashboard is project-level first. It gives you one entrypoint with:

  • aggregate spend and token cards
  • prompt-group and turn rankings
  • model mix and daily trends
  • sessions requiring attention
  • prescriptive insights based on the imported session behavior

Install

pip install harness-observability-layer

This exposes:

hol --help

You can also run the package directly during development:

python -m harness_observability_layer --help

Development Install

For local development from this repository:

pip install -e .

For release validation and packaging tooling during development:

pip install -e .[dev]

Quick Start

From the project you want to inspect:

hol init

That starts a local dashboard server and opens the browser to:

http://localhost:3845

Useful variants:

hol init --no-open
hol init --port 4000
hol init --resolve-files

If you want the aggregate payload that powers the dashboard:

hol data

CLI Surface

HOL currently exposes a deliberately small command surface:

Start the dashboard

hol init

Flags:

  • --port
  • --host
  • --no-open
  • --resolve-files

Print the live aggregate JSON

hol data

Flags:

  • --resolve-files

Local Server

When HOL is running, it serves:

  • / for the dashboard HTML
  • /api/data for the current aggregate JSON
  • /api/refresh to rebuild the in-memory aggregate and return a small status payload

The default bind is:

127.0.0.1:3845

Session Discovery

HOL automatically looks for both Codex and Claude Code archives and filters them to the current project when the archived session records a matching cwd.

Codex discovery

Resolution order:

  1. HOL_CODEX_ARCHIVED_DIR
  2. CODEX_ARCHIVED_SESSIONS_DIR
  3. auto-discovery of local defaults such as:
    • $XDG_DATA_HOME/codex/archived_sessions
    • ~/.config/codex/archived_sessions
    • ~/.codex/archived_sessions

Claude Code discovery

Resolution order:

  1. HOL_CLAUDE_ARCHIVED_DIR
  2. CLAUDE_ARCHIVED_SESSIONS_DIR
  3. auto-discovery of local defaults such as:
    • $XDG_DATA_HOME/claude/projects
    • ~/.config/claude/projects
    • ~/.claude/projects

Sessions without a usable cwd remain eligible. Sessions that clearly belong to another project are skipped.

Privacy And Security

HOL is intended to stay local-first.

Current behavior:

  • session archives are read from local disk
  • normalization and aggregation happen in memory
  • the dashboard is served from localhost
  • HOL does not require a remote backend to produce the dashboard

The localhost dashboard currently exposes project aggregates through /api/data, so it should be treated as local sensitive data while the server is running.

Security-sensitive areas for this repository include:

  • session prompt and tool-output disclosure
  • local path leakage
  • archive discovery and filtering
  • hidden network behavior
  • localhost exposure beyond the intended interface

See also:

Local Development

Run the test suite with:

python -m unittest discover -s tests

If you want to test the local checkout directly:

PYTHONPATH=src python3 -m harness_observability_layer.cli.main --project-root . init --no-open

Repository-Local Helper

This repository still contains lower-level normalization, metrics, and reporting modules that are useful for development and internal testing. The public user-facing product, however, is the localhost dashboard started by hol init.

Citation

If you use this project in research, documentation, or derivative tooling, please cite it using CITATION.cff.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

harness_observability_layer-1.1.1.tar.gz (104.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

harness_observability_layer-1.1.1-py3-none-any.whl (106.7 kB view details)

Uploaded Python 3

File details

Details for the file harness_observability_layer-1.1.1.tar.gz.

File metadata

File hashes

Hashes for harness_observability_layer-1.1.1.tar.gz
Algorithm Hash digest
SHA256 363a3b36a7aaf7d9f8ffd8d33340da30885abd4875c533677550c4f3813553d1
MD5 3b65d3485c773b32bb2554df9e058956
BLAKE2b-256 a983a14874879bc4613656737d8cdd30897d6e024cf1e7e990535229a0418095

See more details on using hashes here.

Provenance

The following attestation bundles were made for harness_observability_layer-1.1.1.tar.gz:

Publisher: release.yml on alexandrelira99/harness_observability_layer

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file harness_observability_layer-1.1.1-py3-none-any.whl.

File metadata

File hashes

Hashes for harness_observability_layer-1.1.1-py3-none-any.whl
Algorithm Hash digest
SHA256 0ada9389fc7a995991789c3ed06f1952d577f5e159be03721cfed688c7bf4821
MD5 b5059c759fc50d4e78a1a193cc64456d
BLAKE2b-256 ee9e6db8d5607526a3a6ec927ee0b05efdc4598f67eb02225c339d4270bcf76c

See more details on using hashes here.

Provenance

The following attestation bundles were made for harness_observability_layer-1.1.1-py3-none-any.whl:

Publisher: release.yml on alexandrelira99/harness_observability_layer

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page