Enhance search quality, intent taxonomy, and MCP functionality

[Szowesgad]

This pull request updates workflow dependencies and corrects organization and product naming throughout documentation and changelogs. The main focus is on pinning GitHub Actions to specific commit SHAs for improved supply chain security and consistency, and standardizing the organization name from "VetCoders" to "Vetcoders" and product references from "memex" to "rust-memex".

Workflow dependency pinning:

• All references to actions/checkout, actions/setup-python, actions/setup-node, actions/upload-artifact, actions/download-artifact, and dtolnay/rust-toolchain in workflow files (.github/workflows/ci.yml, merge-queue-gate.yml, npm-publish.yml, release.yml, retrieval-eval.yml) are now pinned to specific commit SHAs. This improves build reproducibility and mitigates supply chain risks. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]

Organization and product naming consistency:

• The organization name is standardized from "VetCoders" to "Vetcoders" in documentation, changelogs, and markdown files (AGENTS.md, CHANGELOG.md). [1] [2] [3] [4] [5] [6]
• Product references are updated from "memex" to "rust-memex" for clarity and accuracy in changelogs and docs. [1] [2] [3] [4]

Minor corrections and polish:

• Various small doc and changelog typo fixes (e.g., "CodeScribe" → "Codescribe", user path examples, copyright).
  [1] [2] [3] [4]

These changes collectively improve the security, maintainability, and clarity of the repository.

[Copilot]

Copilot wasn't able to review this pull request because it exceeds the maximum number of lines (20,000). Try reducing the number of changed lines and requesting a review from Copilot again.

[gemini-code-assist]

Code Review

This pull request introduces card contract v2, transitioning from legacy bracket headers to YAML frontmatter, and adds a validate-cards command to validate card schemas, headers, hashes, and signal parity. It also implements an index derive command to extract project-scoped semantic indexes, an eval search-quality command to run search quality evaluations, and extends the aicx intents classifier taxonomy to 11 types. Additionally, the MCP server is updated to support streamable HTTP transport with host validation, and the installation scripts are hardened to scan for shadowing aicx-mcp binaries. The review feedback highlights several critical issues: a destructive cleanup bug during dry-run installations, a loose check in source_for_window that incorrectly handles None source paths, ASCII-only character restrictions in search evaluation anchors that strip Unicode/Polish terms, CRLF line-ending incompatibility in card validation, and inconsistent spacing and missing Polish past-tense forms in the outcome line classifier.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

The cleanup_old_binaries calls are executed unconditionally even when AICX_INSTALL_DRY_RUN is set to "1". This means a dry-run installation will still destructively delete existing shadow binaries. Consider wrapping the cleanup calls inside the else block so they are only executed during an actual installation.

[gemini-code-assist]

The source_for_window helper can incorrectly attribute a window to a source path even if some entries in the window have source_path = None. To ensure metadata correctness, the helper should strictly require that every entry in the window shares the same non-empty source path.

fn source_for_window(entries: &[&TimelineEntry]) -> Option<CardSource> {
    let first_entry = entries.first()?;
    let first_path = first_entry.source_path.as_deref()?;
    if entries.iter().any(|entry| entry.source_path.as_deref() != Some(first_path)) {
        return None;
    }

[gemini-code-assist]

Using ch.is_ascii_alphanumeric() in is_anchor_char causes Polish and other non-ASCII alphanumeric characters to be trimmed from query tokens. This can completely destroy or ignore valid search terms (e.g., "łódź" becomes "d" and is ignored due to length constraints). Consider using ch.is_alphanumeric() to support Unicode characters.

Suggested change

	let mut parts = Vec::new();
	push_string(&mut parts, item.get("excerpt"));
	fn is_anchor_char(ch: char) -> bool {
	ch.is_alphanumeric() || matches!(ch, '-' | '_' | '/' | ':' | '.')
	}

[gemini-code-assist]

The check body.starts_with("[signals]\n") will fail to detect the signals block on Windows/CRLF files where the line ending is \r\n. Consider supporting both \n and \r\n line endings to ensure cross-platform compatibility.

Suggested change

	if body.starts_with("[signals]\n")
	&& let Some(end) = body.find("[/signals]")
	if (body.starts_with("[signals]\n") || body.starts_with("[signals]\r\n"))
	&& let Some(end) = body.find("[/signals]")

[gemini-code-assist]

The result verb matching in looks_like_observed_count_outcome_line has inconsistent spacing (e.g., "dal " has no leading space, while " dał " does), which prevents matching "dał" at the start of a line. Additionally, it misses common Polish past-tense forms like "dało" (neuter) and "dały" (plural). Consider aligning the spacing and adding the missing past-tense forms.

Suggested change

	let has_result_verb = lower.contains(" dal ")
	|| lower.contains(" dał ")
	|| lower.contains("dala ")
	|| lower.contains(" dała ")
	|| lower.contains("yielded")
	|| lower.contains("produced")
	|| lower.contains("gave");
	let has_result_verb = lower.contains(" dal ")
	|| lower.contains(" dał ")
	|| lower.contains(" dało ")
	|| lower.contains(" dały ")
	|| lower.contains("dala ")
	|| lower.contains(" dała ")
	|| lower.contains("yielded")
	|| lower.contains("produced")
	|| lower.contains("gave");