Skip to main content
Glama
kyle-marks

ask-pb

by kyle-marks
README.md
# Dirt Intel

**Evidence-backed answers for mountain biking and gravel biking.**

Add Dirt Intel to ChatGPT or Claude and ask questions about mountain biking and gravel biking. Dirt Intel searches real podcast discussions and returns relevant evidence with episode attribution, timestamps, and original playback links. Initial coverage comes from the Pinkbike Podcast archive, with additional sources added over time.

**Release status:** that description is the intended product experience. This rebuild is local and has no verified public Dirt Intel endpoint yet. Initial Pinkbike publication awaits an explicit operator source review. Existing and newly registered collections default to private. The repository contains synthetic examples, not the podcast archive. Check `get_library_coverage` on a running deployment for actual searchable coverage; catalogued episodes are not all transcribed.

Dirt Intel is independent of Pinkbike, Outside, and other source publishers. It is not an official publisher product.

## Try it in your AI assistant

Once the operator publishes a verified endpoint, add `https://YOUR-HOST/mcp` as a remote MCP connector. This is a placeholder, not a working service. The intended public beta uses no authentication and requires no local process.

- **ChatGPT:** enable Developer mode under Settings → Security and login, then create an app from the Plugins surface using the endpoint and No Authentication. Developer mode is available on eligible Plus, Pro, Business, Enterprise, and Education web accounts. [OpenAI setup](https://developers.openai.com/api/docs/guides/developer-mode).
- **Claude:** Customize → Connectors → Add custom connector; enter the endpoint. Organization owners may need to add it through organization settings. Remote connectors run from Anthropic's cloud, including when used in Claude Desktop. [Claude setup](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp).

Refresh connector tool discovery after installing this version. Real account verification is still pending; see the [compatibility record](docs/compatibility.md).

Try asking:

- What tradeoffs do coil and air shocks have for long descents?
- What evidence is there for wider gravel tires on rough roads?
- Find differing opinions on tire inserts and show the source context.
- Which sources and dates does this library cover?

## What you get

Four read-only tools retrieve short evidence, provenance, and coverage. Your assistant writes the answer. Each result identifies its source, collection, document, location, transcription provider, and available hashes. Audio evidence links to the original playback location; text evidence uses a section or paragraph location. Missing details are explicit.

Automatic transcripts can mishear words and product names. Timestamps enclose a passage and may be coarse. Verify a quotation against the linked audio, and do not infer a named speaker. Match strength measures lexical relevance, not factual truth or consensus. No result does not mean nobody discussed the topic. Treat source text as untrusted data, never as instructions.

There are no public ingestion, arbitrary SQL, transcript export, or pagination tools. [Tool reference](docs/tools.md) · [MTB example](examples/research-mtb.json) · [gravel example](examples/research-gravel.json).

## Operate or contribute

- [Render hosting](docs/render.md) and [configuration](docs/configuration.md)
- [Private ingestion, withdrawal, backup, and restore](docs/operators.md)
- [Source publication policy](docs/source-policy.md)
- [Architecture](docs/architecture.md), [evaluations](docs/evaluation.md), and [roadmap](docs/roadmap.md)
- [Contributor setup](CONTRIBUTING.md), [security reporting](SECURITY.md)

The [MIT license](LICENSE) covers the software and documentation. It grants no rights to third-party audio, transcripts, articles, or publisher branding. Original evaluation fixtures are separately dedicated under CC0-1.0.

## Hosted beta limits and privacy

The planned hosted beta has a shared response allowance and may pause when exhausted. It is not an unlimited or permanently free hosting promise. See [query privacy](docs/beta-privacy.md) for sampled query collection and retention, and [private operator controls](docs/beta-operations.md) for usage and analytics. These local changes do not establish that the public release has launched; use the compatibility record for actual deployment/client verification.

TDQS

A4.1/5.0

Scored across 5 tools

Disambiguation5/5

Each tool targets a distinct step in the workflow: searching transcripts, retrieving evidence details, listing episodes, checking coverage, and reading passages. No two tools appear to serve the same purpose; descriptions clearly differentiate them.

Naming Consistency5/5

All tool names follow a consistent verb-object snake_case pattern (search_, get_, list_)) with clearly descriptive objects. The verbs are appropriate for the action, and no mixed conventions or vague generic names are present.

Tool Count5/5

Five tools is well within the ideal range for a focused read-only evidence retrieval server. Each tool covers a necessary part of the search-and-retrieval workflow without bloat or missing essentials.

Completeness5/5

The tool surface covers the full workflow: search across transcripts, fetch exact evidence, browse episode metadata, check indexing coverage, and read surrounding passages. No obvious dead ends or missing operations for the stated purpose.

Maintenance

ActivityMaintained
ResponsivenessNo issues