CodeIssuesDiscussionsWikiPull RequestsProjectsCommitsActionsReleasesContributorsPulse● GatesSecuritySettingsDeploymentsPipelineInsightsAgents✨ Explain✨ Ask AI✨ Workspace✨ Spec✨ Tests▓ Debt Map✨ NL Search🏛 Archaeology
Blame · Line-by-line history
ai-capabilities.json
Each line is annotated with the commit that last touched it. Click any SHA to jump to that commit and see the surrounding change.
| 573a806 | 1 | { |
| 2 | "version": "1.0.0", | |
| 3 | "description": "Complete registry of AI capabilities in Gluecron and how each improves via the flywheel", | |
| 4 | "capabilities": [ | |
| 5 | { | |
| 6 | "name": "Semantic Code Search", | |
| 7 | "description": "Search your codebase by meaning and intent, not just keyword matching", | |
| 8 | "how_it_works": "Embeds code chunks into vector space using Claude's understanding of code semantics. Queries are also embedded and matched by semantic similarity.", | |
| 9 | "model_used": "claude-sonnet-4-6", | |
| 10 | "flywheel_input": "Which results users clicked, whether they found what they needed, subsequent search refinements", | |
| 11 | "flywheel_output": "Improved embedding weighting for your codebase's specific domain language and patterns", | |
| 12 | "competitive_advantage_vs_github": "GitHub search is keyword + regex. Gluecron search understands that 'where do we validate payments' and 'payment validation logic' are the same question." | |
| 13 | }, | |
| 14 | { | |
| 15 | "name": "AI PR Reviewer", | |
| 16 | "description": "Reviews PRs like a senior engineer on your team, not a generic linter", | |
| 17 | "how_it_works": "Reads the full PR diff, surrounding context, and your repo's historical patterns. Posts specific, actionable comments. Learns which comment styles your team accepts.", | |
| 18 | "model_used": "claude-opus-4-7", | |
| 19 | "flywheel_input": "Which review comments were accepted, dismissed, or argued against. What the final merged code looked like.", | |
| 20 | "flywheel_output": "Review style adapts to your team's standards. Comments get more precise and less noisy over time.", | |
| 21 | "competitive_advantage_vs_github": "GitHub Copilot code review is generic. Gluecron's reviewer learns your team's specific standards and enforces them consistently." | |
| 22 | }, | |
| 23 | { | |
| 24 | "name": "Predictive CI", | |
| 25 | "description": "Runs only the tests that matter for the files you changed", | |
| 26 | "how_it_works": "Analyses changed files, maps them to test coverage data and historical failure correlations. Selects the minimal test set most likely to catch regressions.", | |
| 27 | "model_used": "claude-haiku-4-5", | |
| 28 | "flywheel_input": "Whether the selected tests caught the failures, whether missed tests would have caught additional failures", | |
| 29 | "flywheel_output": "Improved file-to-test mapping accuracy. Faster CI with higher confidence.", | |
| 30 | "competitive_advantage_vs_github": "GitHub Actions runs everything every time. Gluecron predicts which tests matter and cuts CI time by up to 60% without missing failures." | |
| 31 | }, | |
| 32 | { | |
| 33 | "name": "Smart Merge", | |
| 34 | "description": "Resolves merge conflicts by understanding semantic intent, not just line differences", | |
| 35 | "how_it_works": "When a conflict occurs, reads both branches' changes in context, understands the intent of each, and proposes a resolution that satisfies both.", | |
| 36 | "model_used": "claude-opus-4-7", | |
| 37 | "flywheel_input": "Whether the proposed resolution was accepted, modified, or rejected. What the final resolution looked like.", | |
| 38 | "flywheel_output": "Improved conflict resolution accuracy for your codebase's common patterns.", | |
| 39 | "competitive_advantage_vs_github": "GitHub shows you the conflict. Gluecron resolves it." | |
| 40 | }, | |
| 41 | { | |
| 42 | "name": "Auto Documentation", | |
| 43 | "description": "Generates and keeps documentation in sync with code changes", | |
| 44 | "how_it_works": "On every PR merge, scans changed files, updates or generates corresponding documentation, opens a follow-up PR with doc updates.", | |
| 45 | "model_used": "claude-sonnet-4-6", | |
| 46 | "flywheel_input": "Whether generated docs were accepted as-is, edited, or rejected. What the final docs looked like.", | |
| 47 | "flywheel_output": "Documentation style matches your team's voice and technical depth preferences.", | |
| 48 | "competitive_advantage_vs_github": "GitHub has no documentation automation. Gluecron ensures docs never go stale." | |
| 49 | }, | |
| 50 | { | |
| 51 | "name": "Security Intelligence", | |
| 52 | "description": "Explains vulnerabilities in plain language with prioritized, context-aware fix suggestions", | |
| 53 | "how_it_works": "Scans dependencies and code on every push. For each finding, generates an explanation of the risk in your specific context and a concrete fix suggestion.", | |
| 54 | "model_used": "claude-sonnet-4-6", | |
| 55 | "flywheel_input": "Which findings developers fixed first, which they deferred, whether AI-suggested fixes were applied", | |
| 56 | "flywheel_output": "Improved prioritization: surfaces the vulnerabilities your team is most likely to act on, in the order most impactful for your risk profile.", | |
| 57 | "competitive_advantage_vs_github": "GitHub Dependabot lists CVEs. Gluecron explains what they mean for your code and tells you exactly what to change." | |
| 58 | }, | |
| 59 | { | |
| 60 | "name": "Codebase Q&A", | |
| 61 | "description": "Ask questions about the repo in natural language, get accurate answers", | |
| 62 | "how_it_works": "Maintains a live semantic index of the codebase. Answers questions by retrieving relevant code chunks and synthesizing an accurate response.", | |
| 63 | "model_used": "claude-sonnet-4-6", | |
| 64 | "flywheel_input": "Whether the answer was rated helpful, follow-up questions asked, source code references clicked", | |
| 65 | "flywheel_output": "More accurate retrieval and better answer synthesis for your specific codebase structure.", | |
| 66 | "competitive_advantage_vs_github": "GitHub has no Q&A. Gluecron is the institutional knowledge layer for your codebase." | |
| 67 | }, | |
| 68 | { | |
| 69 | "name": "Release Note Generator", | |
| 70 | "description": "Writes release notes that match your team's style and voice", | |
| 71 | "how_it_works": "Reads merged PRs and commits since last release, synthesizes a release note in your team's established style. Learns from edits to previous generated notes.", | |
| 72 | "model_used": "claude-sonnet-4-6", | |
| 73 | "flywheel_input": "How much the generated notes were edited before publishing", | |
| 74 | "flywheel_output": "Style drift toward zero edits needed over time.", | |
| 75 | "competitive_advantage_vs_github": "GitHub has no release note AI. Teams spend hours on this manually." | |
| 76 | } | |
| 77 | ] | |
| 78 | } |