qdash-core

name	qdash-core
description	Connect agents to a running QDash instance through qdash-client and saved local profiles. Use when an agent needs QDash profile discovery, safe authentication handling, project context, read-only API helpers, raw read-only endpoint access, or shared QDash access rules before using more specific QDash skills.

QDash Core

Overview

Use this skill to let an agent query QDash through qdash-client instead of scraping the UI, calling curl directly, or hand-writing auth headers. Prefer saved profiles for credentials and keep tokens, Cloudflare Access secrets, passwords, and raw config contents out of conversation output. This is the common access layer for the other QDash user-facing skills.

For public/general use, assume qdash-client is the required API transport. Do not install it permanently unless the user asks; prefer ephemeral execution with uv run --with qdash-client.

Quick Start

1. Locate the profile without printing secrets:

uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py config-path

2. Run a read-only smoke check:

uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local chips

3. Query common API data:

uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local default-chip
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local metrics-config
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local chip-metrics --chip-id chip-001
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local chip-qubits --chip-id chip-001 --limit 20
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local task-results --limit 20 --status success
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local task-result --task-id task-001
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local task-knowledge --task-name t1
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local qubit-latest --task t1
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local issues --limit 20
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local flows
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local flow-templates
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local files-tree
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local ai-reviews --limit 20
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local provenance-stats
uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py --profile local timeseries --parameter t1 --start-at 2026-06-01T00:00:00Z --end-at 2026-06-08T00:00:00Z

For QDash development, set QDASH_REPO_PATH to a local oqtopus-team/qdash checkout to test unreleased client changes. The helper also probes ~/src/github.com/oqtopus-team/qdash.

If using a local checkout and plain python does not have client dependencies, run the helper through uv:

uv run --with httpx --with pydantic python skills/qdash/qdash-core/scripts/qdash_query.py --profile local chips

Workflow

1. Use config-path first when the user mentions a local profile. Report only the path, available profile names, and whether the file exists.
2. Prefer uv run --with qdash-client python skills/qdash/qdash-core/scripts/qdash_query.py ... for supported read-only operations because it handles profile loading, auth headers, project headers, model serialization, and error formatting through the official client.
3. Use named read-only commands first: chips, default-chip, metrics-config, chip-metrics, chip-qubits, chip-qubit, chip-couplings, chip-coupling, timeseries, task-results, task-result, task-note, task-result-issues, task-knowledge, task-knowledge-markdown, qubit-latest, qubit-history, coupling-latest, coupling-history, projects, project, files-tree, file-content, git-status, issues, issue-knowledge, flows, flow, flow-templates, flow-template, flow-helper-files, flow-helper-file, executions, ai-reviews, ai-review-runs, ai-review-run, and provenance-*.
4. Use raw-get for read-only endpoints not covered by first-class commands. It still goes through qdash-client; do not use curl as the normal path.
5. Use a local checkout fallback only for qdash-client development or unreleased API/client changes.
6. Before creating, updating, deleting, excluding, re-executing, pushing files, pulling files, or triggering flows, state the exact endpoint/action and wait for user confirmation.
7. Summarize returned data instead of dumping huge JSON. Include counts, IDs, time ranges, and suspicious values that answer the user's question.

API Reference

Read references/qdash.md when you need endpoint groups, local source locations, or examples for extending the helper.

Primary local sources in the QDash checkout:

• src/qdash/client/README.md
• docs/user-guide/qdash-client.md
• docs/oas/openapi.json
• src/qdash/client/services/client.py

Safety

• Never print api_token, password, cf_access_client_secret, or full profile contents.
• Do not reconstruct auth headers in shell or curl unless the user explicitly asks for low-level debugging.
• Treat /files/git/push, /files/git/pull, /flows/*/execute, re-execute, exclude, admin, auth, and project membership endpoints as write or operationally sensitive.
• Prefer absolute UTC ISO timestamps with trailing Z for time-series calls.
• Close QDashClient instances when writing custom snippets.