Skip to content
PyCharter plugin · 0.1.0

Check a batch from your coding agent.

Ask Codex to validate local records against a versioned contract and save the actual result. Start with the synthetic price example, then use the same workflow for your own JSON batch.

Tested with Codex CLI 0.162.1, Python 3.13 and pycharter 0.0.67. No hosted account, FMP key, MCP server or background process is required.

01 / INSTALL

Add the local plugin.

Download and extract the bundle, then run these commands from its directory. Keep that directory available for updates. The bundle contains the skill, pinned example, local helpers and verification records.

unzip pycharter-0.1.0.zip -d optophi-plugin
cd optophi-plugin/optophi-pycharter-0.1.0
codex plugin marketplace add .
codex plugin add pycharter@optophi-local

Installation adds the plugin to Codex. Python dependencies are a separate setup step; the skill uses an isolated environment or a compatible existing interpreter.

02 / ASK

Start with a fresh local result.

Start a new Codex session in the project where you want the files, then ask:

Use $pycharter-validate to create the offline price-batch example here and reproduce its actual validation evidence. Preserve all source files and report where the fresh evidence is saved.

The agent creates the example, runs the published PyCharter validator, and saves fresh evidence. If a prerequisite is missing, it reports the blocked check. The browser walkthrough remains a recorded example.

03 / INSPECT

Keep the rule, input and failure together.

The example contains a contract, valid and rejected batches, a known synthetic correction, and actual validation results. One of two rejected-batch records fails because its close is −8.25.

  • Exact inputsEvidence names the input and contract hashes, runtime version, check time, counts and diagnostics.
  • A legitimate correctionThe synthetic reference supplies 8.25. Your own rejected value needs a trusted source; an unknown value stays unresolved.
  • A fresh checkSave new evidence for each check and retain earlier failures. A remembered conversation is insufficient evidence.
04 / REUSE

Use your own batch, or return tomorrow.

For local self-contained JSON contracts and JSON arrays of records, a natural request can select the installed skill:

Validate batch.json against contract.json before I load this batch. Save actual diagnostics and file hashes to a new evidence file. Preserve the input files; there is no trusted source for correcting rejected values.

If the agent does not select it, explicitly name $pycharter-validate. To resume saved work:

Inspect the saved input and contract, rerun the rejection check, and save fresh evidence to a new file. Preserve all previous files and report whether the result is a newly executed check.
05 / MANAGE

Update or remove it on your terms.

Update the extracted source, reinstall with codex plugin add pycharter@optophi-local, and start a new session. To disable without removing, set enabled = false under [plugins."pycharter@optophi-local"] in your applicable Codex configuration.

codex plugin remove pycharter@optophi-local
codex plugin marketplace remove optophi-local

Removing the plugin leaves your project files and evidence available for direct Python use.

Verified scope: local record validation in Codex. Claude Code installation compatibility is included; its agent behavior is unverified. The plugin provides task guidance and maintained validation. Database publication, FMP extraction, schedules, freshness monitoring and control of other agent tools have separate requirements.

Inspect the bundled README and verification records for tested behavior and limits. This historical 0.1.0 download keeps its original bytes and notices; it is separate from the unpublished pipeline plugin.

Download fingerprints and source identities. Current CLI/plugin development belongs to Optophi platform tooling; its source acceptance does not qualify a public installer or model session.

Explore the FMP pipeline direction
Feedback