Resolution Capsule

AI coding assistants are trained on Stack Overflow. When they fix your bug, that fix should go back to Stack Overflow.

Resolution Capsule closes the loop: it captures an AI-generated fix, strips everything private (tokens, emails, internal paths, IPs, ticket IDs), and posts a clean, reviewable draft directly to Stack Overflow.

---

The Problem

Every day, AI assistants help developers fix bugs using knowledge that was scraped from public forums. Those fixes stay in a chat window and never benefit the next person who hits the same error. The knowledge loop is broken.

What It Does

1. You fix a bug — with an AI assistant or manually
2. Resolution Capsule captures the fix and its context
3. It redacts secrets, identity markers, and internal details automatically
4. It produces a Stack Overflow-ready draft and posts it via the API

The goal is zero friction between "fix applied" and "community helped."

---

Surfaces

Surface	Best for
VS Code extension	Capturing AI-generated fixes directly from your editor (git diff → post)
CLI	CI pipelines, scripts, batch processing of resolved incidents
Web UI	Quick one-off drafts and testing the sanitizer
AI skill	Embedding the workflow inside Cursor, Copilot, Claude Code, or any agent

---

VS Code Extension

Install from the marketplace or from the .vsix in vscode-extension/.

Key command — Create from AI Fix:

1. Use your AI assistant to apply a fix
2. Cmd+Shift+P → Resolution Capsule: Create from AI Fix
3. Describe the problem and root cause
4. A sanitized draft opens — one click posts it to Stack Overflow

The command captures your current git diff as the fix automatically. No copy-pasting.

Other commands: Create From Selection, Create From Prompts, Connect Stack Overflow, Disconnect Stack Overflow.

---

CLI

python3 capsule_cli.py \
  --problem "Build fails after dependency upgrade" \
  --rootCause "Lockfile drift" \
  --fix "Regenerated package-lock.json" \
  --source "ai-assisted"

JSON input:

python3 capsule_cli.py --format json --json '{"problem":"...","rootCause":"...","fix":"...","source":"ai-assisted","mode":"balanced"}'

Write to file:

python3 capsule_cli.py --output capsule.md --problem "..." --rootCause "..." --fix "..."

---

Web UI

python3 app.py

Open http://127.0.0.1:8000.

---

Sanitization

Balanced mode (default): API keys, bearer tokens, JWTs, GitHub tokens, Slack tokens, AWS keys, emails, IPs, internal URLs, filesystem user paths, private key blocks.

Strict mode: everything above plus ticket IDs and long numeric identifiers.

---

What's Missing (and Where to Help)

This is a work in progress. The pieces that matter most and aren't built yet:

• Cursor / Copilot / Claude Code hooks — trigger a capsule automatically when an AI fix is accepted, without the developer doing anything manually
• Entropy-based secret detection — catch high-entropy strings that don't match known patterns
• Duplicate detection — check if the question already exists on Stack Overflow before posting
• Team policy packs — org-level redaction rules layered on top of the defaults
• JetBrains plugin — same workflow as the VS Code extension

If any of these interest you, the codebase is small and the engine is pure Python with no dependencies.

---

Stack Overflow Setup

To enable auto-posting, register a free app at stackapps.com/apps/oauth/register with OAuth domain localhost, then add your credentials to VS Code settings:

resolutionCapsule.stackOverflow.clientId
resolutionCapsule.stackOverflow.clientSecret
resolutionCapsule.stackOverflow.apiKey

Run Resolution Capsule: Connect Stack Overflow once to authorise.

---

Contributing

The repo is at github.com/adivarshney/resolution-capsule. Issues, PRs, and ideas are all welcome. If you're unsure where to start, the AI tool integrations section above is the frontier.

---

Run the Tests

python3 -m unittest test_app -v