RepoMind 2.0.0b11

RepoMind 2.0.0b11 is a hardening beta focused on repository trust boundaries and release validation.

Highlights

• Hardened RepoMind index/cache containment against symlink, Windows reparse-point, and hardlink redirection.
• Parser recursion and resource failures are isolated per file so incomplete results are discarded without aborting the full index.
• Passive Git inspection is hardened against configured executable behavior such as fsmonitor, hooks, external diff, and textconv.
• Ordinary Git semantics are preserved, including Windows CRLF and core.autocrlf behavior.
• Added Linux and Windows CI plus installed-wheel release gates.
• Verified the v2.0.0b10 to v2.0.0b11 index and manual-memory lifecycle.

Installation and compatibility

• The wheel attached to this GitHub Release is the authoritative distribution. Do not use plain pip install repomind; that name may resolve to another project on PyPI.
• Python 3.11-3.13 are explicitly release-tested for b11. Other Python versions are not claimed as release-tested.
• Windows and Linux installed-wheel checks passed. macOS is not claimed as release-tested for this beta.
• The b10 index and manual-memory upgrade path was tested against the b11 candidate.
• Package uninstall preserves repository index, configuration, and manual memory.
• repomind init --force rebuilds the database and removes database-only manual memory; export or preserve needed data first.

Security / reliability notes

• RepoMind performs static analysis and does not execute recommended repository test commands; those commands remain recommendations/data only.
• Passive Git execution paths were narrowed to prevent configured helper programs from running during inspection.
• Parser resource failures are isolated, but RepoMind is not an operating-system sandbox.
• Filesystem preflight validation reduces redirection risk but does not eliminate time-of-check/time-of-use (TOCTOU) races.
• Downstream MCP clients or agents may send requested metadata or snippets to their configured provider.
• This beta is not a security certification, does not claim every vulnerability is fixed, and does not claim exhaustive test-impact coverage.

Artifact SHA-256

• repomind-2.0.0b11-py3-none-any.whl: 3202f593084a5f13cbe68c21bc7a2bc3c6207671a28e10303267bf0b0594ba9f
• repomind-2.0.0b11.tar.gz: 8814e9ab9ca134fc7b34ae8b1f91030490411a29c39c8af5c660bb8dacb24938

RepoMind 2.0.0b10

RepoMind 2.0.0b10 is a public beta focused on bounded, explainable repository analysis and change-aware test selection.

Highlights

• Added bounded test-impact analysis through the CLI and MCP, including changed files, impacted areas, relevant existing tests, uncovered areas, recommended commands, evidence, confidence, and completeness metadata.
• Bounded Git changed-file discovery and evidence retention during processing while preserving deterministic, priority-aware selection.
• Added deterministic Python and JavaScript/TypeScript test ownership and command recommendations, including route, API, migration, and cross-stack evidence.
• Hardened Git parsing, path validation, unusual filenames, renames, stream chunk boundaries, timeouts, diagnostics, large change sets, and tight evidence budgets.
• Kept test-impact analysis-only: RepoMind returns test commands as data and never executes repository code or tests.

Installation and compatibility

• Added verified Windows wheel installation, local MCP setup, upgrade, uninstall, and reinstall guidance.
• A schema-2 index created by v2.0.0b9 remained readable in the tested b10 upgrade path without reindexing. This is not a guarantee for every future schema.
• Corrected package/runtime version metadata for 2.0.0b10.
• Included the complete Apache License 2.0 terms in the wheel, source distribution, and repository. This corrects the packaged license terms in b10; historical b9 artifacts remain unchanged.

Important limitations

• Static analysis can miss runtime behavior, dynamic dispatch, generated code, framework behavior, and external services.
• Test recommendations are evidence-based starting points, not exhaustive coverage guarantees.
• Confidence labels summarize static evidence and are not calibrated probabilities.
• Respect truncation, completeness, and lower-bound metadata when interpreting counts.
• RepoMind is not a security certification, and agent conclusions still require developer review.

Installation

Download the wheel attached to this release and install it in an isolated environment. Do not use plain pip install repomind; that name on PyPI currently resolves to a different project.

Artifact SHA-256

• repomind-2.0.0b10-py3-none-any.whl: 8ac3533c831a9aae1032081c3a7083761d8e3aed09a69f6cc92293cf6405f9b1
• repomind-2.0.0b10.tar.gz: f026f04e518843048190fcd4d43dbd089aaf9f6f876fb4bf6a0e863961bfd920

RepoMind 2.0.0b9 (Beta)

RepoMind 2.0.0b9

RepoMind 2.0.0b9 is a beta release focused on more accurate code-context retrieval across Python, TypeScript, and React repositories.

Improvements

• Suppresses .agents/skills/** noise during normal application-code queries.
• Preserves skill retrieval when a query explicitly requests skill information.
• Improves retrieval for related concepts such as stock universe, symbols, tickers, watchlists, and scanner inputs.
• Strengthens bounded dependency and import expansion.
• Adds lexical content fallback for conceptual repomind snippets queries.
• Improves bounded TypeScript/React traversal across API clients, hooks, models, pages, and components.
• Adds realistic finance/full-stack retrieval regression coverage.
• Prevents workflow tests from modifying tracked fixture files.

Validation

• 113 automated tests passed.
• Ruff passed.
• Mypy passed.
• Package build and clean-wheel installation passed.
• Retrieval and weighted evaluation benchmarks passed.
• Verified against a real 171-file Python/React/TypeScript F&O analysis project with zero parse errors.
• Verified direct installation from the GitHub tag in an isolated environment.

Installation

Install the wheel attached to this release:

python -m pip install https://github.com/raazlehra/Repomind/releases/download/v2.0.0b9/repomind-2.0.0b9-py3-none-any.whl

RepoMind v2.0.0-beta.5

RepoMind v2.0.0-beta.5 adds the beta repository audit workflow and keeps generated local audit reports out of the repository.

Highlights:

• New repomind audit command
• Deterministic Markdown and JSON audit reports
• Architecture, framework, route, test, and important-file detection
• Static release-risk findings
• False-positive reductions for test-only security evidence and validated parent provisioning flows
• Polished paid-audit-style report format
• Sanitized sample audit report in docs/sample-audit-report.md
• Generated local audit reports ignored and removed from Git tracking

Validated with:

• python -m pytest -q
• python -m ruff check .
• python -m mypy repomind
• python -m build
• python -m tools.evaluate_weighted
• python -m tools.evaluate_retrieval
• python -m benchmarks.run_benchmark --format markdown

Real-repo validation:

• School Management System
• Invoice Payment Reminder App

Note:
This is a beta release. Audit findings are static heuristics intended to guide review; they are not a security certification.

RepoMind v2.0.0-beta.3

RepoMind v2.0.0-beta.3 adds the beta repository audit workflow.

Highlights:

• New repomind audit command
• Deterministic Markdown and JSON audit reports
• Architecture, framework, route, test, and important-file detection
• Static release-risk findings
• False-positive reductions for test-only security evidence and validated parent provisioning flows
• Polished paid-audit-style report format
• Sanitized sample audit report in docs/sample-audit-report.md
• Generated local audit reports ignored by Git

Validated with:

• python -m pytest -q
• python -m ruff check .
• python -m mypy repomind
• python -m build
• python -m tools.evaluate_weighted
• python -m tools.evaluate_retrieval
• python -m benchmarks.run_benchmark --format markdown

Real-repo validation:

• School Management System
• Invoice Payment Reminder App

Note:
This is a beta release. Audit findings are static heuristics intended to guide review; they are not a security certification.