Skip to main content

Turn complex code into connected LLM context, business workflows, and critical risk findings

Project description

UnvibeCode

Vibe the prototype. Unvibe the codebase.

Turn complex vibe code into verified engineering—understand its connections, business workflows, and critical business risks.

Vibe coding works well for prototypes. As a repository grows, behavior becomes spread across files, services, state changes, and external calls. UnvibeCode reviews that repository as one connected system and produces customer-ready HTML reports plus reusable LLM context.

Install and review in two commands

pip install unvibecode
python -m unvibecode review --repository "/path/to/repository"

Windows example:

python -m unvibecode review --repository "D:\upstox"

macOS or Linux users can use python3 when that is their installed Python command:

python3 -m unvibecode review --repository "/Users/example/repository"

That is the complete customer setup:

  • No activation or access key.
  • No OpenAI API key or .env file.
  • No virtual environment, wheel, ZIP, or PATH configuration.
  • The repository path is the only required input.
  • Results save automatically to ./shipready_results.

What UnvibeCode produces

1. Connected Code Map for LLMs

Large repositories are difficult to understand one file at a time. The interactive code map shows meaningful file, symbol, import, and call connections. Select a file to download a compact package containing that file and its relevant connected code, ready to give to an LLM.

Connected Code Map showing contextually connected files and an LLM-ready download

2. Complete Repository Context

Repeatedly rebuilding repository context for every LLM conversation is slow and inconsistent. UnvibeCode creates one normalized ZIP containing ordered code chunks, verified connections, context selections, metadata, and usage instructions for later LLM or API workflows.

complete_repository_context_for_llm.zip
├── manifest.json
├── chunks.jsonl
├── connections.jsonl
├── selections.jsonl
└── README.md

3. Business Workflow Map

A technical file tree does not explain what the software does for customers or the business. UnvibeCode reconstructs related code paths as understandable workflows, showing what starts a workflow, the important decisions and state changes, and the resulting business outcome.

Illustrative workflow connecting checkout, payment processing, and fulfilment

4. Business Risk Findings

Technical scanners can miss failures that emerge across a connected business workflow. UnvibeCode identifies critical business workflow risks in the code and shows the supporting evidence. Each finding explains the trigger, what the code does, the business rule, the downside, affected customers or assets, what to change, and how to verify the correction.

Illustrative payment risk with business path, remediation, acceptance check, and code evidence

If no candidate passes the evidence threshold, the report clearly records a no-findings outcome for the completed review scope.

The images are illustrative examples. Workflows and findings are generated from the repository being reviewed.

Customer results

For a supported repository, the automatically created results folder contains:

shipready_results/
└── analysis_<timestamp>/
    └── customer_results/
        ├── 01_connected_code_map_for_llm.html
        ├── complete_repository_context_for_llm.zip
        ├── 02_business_workflow_map.html
        └── 03_business_risk_findings.html

The terminal displays continuous progress. When the review completes, the HTML reports open automatically.

Large repositories

Repositories above approximately two million estimated source tokens receive the Connected Code Map and Complete Repository Context. The deeper business workflow and risk review does not run above that threshold.

Initial public observation period

During the initial two-day public observation period, UnvibeCode requires no activation, access key, customer API key, usage charge, or customer quota. Technical request-size and concurrency protections remain in place to keep the service reliable. Usage controls may be introduced in a later package version after actual demand is understood.

Data processing

Repository scanning, graph construction, connected-context preparation, and report rendering run on the customer's computer. Relevant structured code context is sent securely to the hosted UnvibeCode analysis service for the business workflow and risk review. Customers do not provide an OpenAI key.

Review only repositories you are authorized to process.

Requirements

  • Python 3.11 or newer.
  • Internet access for the hosted business analysis.
  • Read access to the repository being reviewed.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

unvibecode-0.3.1.tar.gz (314.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

unvibecode-0.3.1-py3-none-any.whl (321.1 kB view details)

Uploaded Python 3

File details

Details for the file unvibecode-0.3.1.tar.gz.

File metadata

  • Download URL: unvibecode-0.3.1.tar.gz
  • Upload date:
  • Size: 314.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/7.0.0 CPython/3.13.3

File hashes

Hashes for unvibecode-0.3.1.tar.gz
Algorithm Hash digest
SHA256 746930081809b6d104773b3e0af5d08dcbc1811943a83a55094fc8266b602385
MD5 296e08afbc05bdac2f4dbc380ac3f270
BLAKE2b-256 cb2f0a8516eaf35915eb9c4e23bc2805a208411b3f41f0e962a5ca64ad4ea396

See more details on using hashes here.

File details

Details for the file unvibecode-0.3.1-py3-none-any.whl.

File metadata

  • Download URL: unvibecode-0.3.1-py3-none-any.whl
  • Upload date:
  • Size: 321.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/7.0.0 CPython/3.13.3

File hashes

Hashes for unvibecode-0.3.1-py3-none-any.whl
Algorithm Hash digest
SHA256 a63cf785e051df90545846c25f7660dfc4b47edf07e196a058ccb0e31b35edf5
MD5 81cab6c6551ac7b9e285d2c41cdb92ef
BLAKE2b-256 d46f44972396a86a3dd63d600aa7014eac5609ee6008a0aa79a7f7e64e119679

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page