Commit573a806
feat: establish Gluecron as standalone GitHub-superseding platform with AI flywheel
4 files changed+376−0573a806640a9946af843909f044e23e97996d556
4 changed files+376−0
Added.ai-flywheel/ai-capabilities.json+78−0View fileUnifiedSplit
@@ -0,0 +1,78 @@
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}
Added.ai-flywheel/config.json+116−0View fileUnifiedSplit
@@ -0,0 +1,116 @@
1{
2 "version": "1.0.0",
3 "product": "gluecron",
4 "description": "AI flywheel configuration for the Gluecron code platform",
5
6 "signal_collection": {
7 "pull_requests": [
8 "pr_id",
9 "repo",
10 "author",
11 "files_changed",
12 "lines_added",
13 "lines_removed",
14 "review_comments_by_ai",
15 "review_comments_accepted",
16 "review_comments_dismissed",
17 "time_to_first_review_ms",
18 "time_to_merge_ms",
19 "ci_pass_on_first_run",
20 "reviewer_assigned_by_ai",
21 "reviewer_assignment_accepted",
22 "merge_conflict_occurred",
23 "merge_conflict_resolved_by_ai"
24 ],
25 "ci_runs": [
26 "run_id",
27 "repo",
28 "branch",
29 "commit_sha",
30 "tests_selected_by_ai",
31 "tests_selected_total",
32 "total_tests_in_suite",
33 "failures_caught_in_selected",
34 "failures_missed",
35 "duration_ms",
36 "duration_savings_ms"
37 ],
38 "code_search": [
39 "query",
40 "repo",
41 "result_count",
42 "result_clicked",
43 "result_position_clicked",
44 "user_found_what_they_needed"
45 ],
46 "security_scans": [
47 "finding_id",
48 "severity",
49 "cve_id",
50 "file_path",
51 "ai_explanation_rating",
52 "fix_applied",
53 "time_to_fix_hours",
54 "fix_correct"
55 ]
56 },
57
58 "learning_cycles": {
59 "on_pr_review_comment_resolved": {
60 "action": "update_review_style_model_for_repo",
61 "model": "claude-sonnet-4-6",
62 "latency_budget_ms": 1000
63 },
64 "on_pr_merge": {
65 "action": "score_prediction_accuracy_and_update_file_risk_index",
66 "model": "claude-sonnet-4-6",
67 "latency_budget_ms": 2000
68 },
69 "on_ci_complete": {
70 "action": "update_test_selection_model_with_coverage_outcome",
71 "model": "claude-haiku-4-5",
72 "latency_budget_ms": 500
73 },
74 "on_issue_close": {
75 "action": "update_issue_classification_and_resolution_pattern_index",
76 "model": "claude-haiku-4-5",
77 "latency_budget_ms": 500
78 },
79 "daily_repo_health": {
80 "action": "generate_repo_health_report_per_active_repo",
81 "model": "claude-sonnet-4-6",
82 "schedule": "0 8 * * *",
83 "output": "repo_dashboard_and_owner_digest_email"
84 }
85 },
86
87 "improvement_targets": [
88 {
89 "metric": "pr_review_acceptance_rate",
90 "measure": "ai_comments_accepted / ai_comments_posted",
91 "target": 0.70
92 },
93 {
94 "metric": "ci_test_selection_coverage",
95 "measure": "failures_caught_in_selected / total_failures",
96 "target": 0.95,
97 "with_test_reduction_target": 0.40
98 },
99 {
100 "metric": "reviewer_assignment_accuracy",
101 "measure": "pct_of_ai_assigned_reviewers_who_complete_review",
102 "target": 0.85
103 },
104 {
105 "metric": "security_fix_acceptance_rate",
106 "measure": "pct_of_ai_suggested_fixes_applied",
107 "target": 0.75
108 }
109 ],
110
111 "data_retention": {
112 "per_repo_rolling_months": 12,
113 "anonymized_aggregate_months": 36,
114 "delete_on_repo_delete": true
115 }
116}
AddedPRODUCT_MANIFEST.json+80−0View fileUnifiedSplit
@@ -0,0 +1,80 @@
1{
2 "product": "Gluecron",
3 "tagline": "The AI-native code platform. Every commit gets smarter.",
4 "version": "1.0.0",
5 "type": "standalone-platform",
6 "category": "developer-platform",
7 "purpose": "AI-native code hosting, review, and CI/CD platform that supersedes GitHub",
8 "standalone": true,
9 "supersedes": "GitHub",
10 "also_replaces": ["GitLab", "Bitbucket", "Gitea", "Forgejo"],
11 "core_capabilities": [
12 "git_hosting",
13 "ai_code_review",
14 "intelligent_ci_cd",
15 "issue_tracking",
16 "project_management",
17 "dependency_analysis",
18 "security_scanning",
19 "vscode_extension",
20 "cli_tools",
21 "api_access"
22 ],
23 "ai_advantages_over_github": [
24 "Semantic code search that understands meaning, not just keyword matching",
25 "AI PR reviewer that thinks like a senior engineer and learns from accepted/rejected comments",
26 "Predictive CI that knows which tests to run based on semantic analysis of changed files",
27 "Smart merge conflict resolution that understands semantic intent, not just line diffs",
28 "Auto-generated and maintained documentation that stays in sync with code changes",
29 "Security intelligence that explains vulnerabilities in plain language with prioritized fix suggestions",
30 "Codebase Q&A: ask questions in natural language, get answers that improve over time",
31 "AI-written release notes that learn your team's style and voice"
32 ],
33 "stack": {
34 "runtime": "Bun v1.x",
35 "database": "PostgreSQL 16 via Drizzle ORM",
36 "deployment": "Docker on Crontech bare metal",
37 "cli": "gc (gluecron CLI, installable via bun or curl)",
38 "ide_integration": "VSCode extension"
39 },
40 "integrates_with_via_api_only": {
41 "crontech": {
42 "relationship": "Gluecron is HOSTED on Crontech",
43 "push_trigger": "POST /api/v1/deploy on every merge to main"
44 },
45 "gatetest": {
46 "webhook": "POST /api/gate/result",
47 "purpose": "Block PR merges that fail test gates"
48 },
49 "alecrae": {
50 "relationship": "AlecRae repos are hosted on Gluecron"
51 }
52 },
53 "ai_flywheel": {
54 "enabled": true,
55 "core_principle": "Every interaction (PR review, CI run, issue close, code search query) feeds back to make the next interaction smarter",
56 "models": {
57 "deep_code_analysis": "claude-opus-4-7",
58 "standard_analysis": "claude-sonnet-4-6",
59 "real_time_autocomplete": "claude-haiku-4-5"
60 },
61 "learns_from": [
62 "pr_review_comment_acceptance_rates",
63 "ci_pass_rates_by_file_path",
64 "issue_resolution_time_by_label",
65 "code_churn_by_module",
66 "merge_conflict_frequency_by_file",
67 "test_coverage_trends",
68 "security_finding_fix_prioritization",
69 "code_search_query_satisfaction"
70 ],
71 "improves": [
72 "pr_review_suggestion_relevance",
73 "ci_test_selection_efficiency",
74 "reviewer_assignment_accuracy",
75 "bug_prediction_by_file_churn",
76 "security_fix_prioritization"
77 ]
78 },
79 "competitive_position": "GitHub Copilot is an add-on. Gluecron is AI from the ground up. Every feature, every workflow, every data model is designed for AI-native operation. The AI that reviews your PR today is smarter than it was yesterday, because yesterday's accepted and rejected suggestions trained it. GitHub cannot replicate this without rebuilding from scratch."
80}
Addedcli/STANDALONE_CLI.md+102−0View fileUnifiedSplit
@@ -0,0 +1,102 @@
1# Gluecron CLI (`gc`)
2
3The `gc` CLI replaces `git` + `gh` + GitHub Actions CLI with a single tool that understands your entire development workflow.
4
5## Installation
6
7```bash
8# Via Bun
9bun install -g gluecron
10
11# Via install script
12curl -fsSL https://get.gluecron.com | sh
13```
14
15## Configuration
16
17```bash
18export GLUECRON_URL=https://your-gluecron-instance.com
19export GLUECRON_TOKEN=gc_your_api_token
20
21# Or run once:
22gc auth login
23```
24
25## Core Commands
26
27### Repository
28```bash
29gc init # Initialise a new Gluecron repo
30gc clone owner/repo # Clone a repo
31gc push # Push changes (like git push, but smarter)
32gc push --draft # Push and open a draft PR
33```
34
35### Pull Requests
36```bash
37gc pr create # Open a PR (AI generates title + description)
38gc pr create --ai-description # AI writes full PR description from diff
39gc pr review # Get AI review of the current branch
40gc pr review --file src/auth.ts # AI review focused on a specific file
41gc pr merge # Merge current PR
42gc pr list # List open PRs
43```
44
45### CI
46```bash
47gc ci status # Show CI status for current branch
48gc ci logs # Stream CI logs
49gc ci rerun # Rerun failed CI jobs
50```
51
52### Deployment
53```bash
54gc deploy # Trigger deploy for current branch
55gc deploy --env production # Deploy to specific environment
56gc logs --env production # Stream production logs
57gc rollback # Rollback last deployment
58```
59
60### AI Features
61```bash
62gc ai ask "why is the auth test failing" # Ask about the codebase
63gc ai ask "where do we handle rate limiting" # Semantic code search via CLI
64gc ai review # Full AI code review
65gc ai fix # AI suggests fixes for current errors
66gc ai docs # Generate docs for changed files
67```
68
69### Issues
70```bash
71gc issue create # Create issue (AI suggests labels/assignees)
72gc issue list # List issues
73gc issue close 42 # Close issue
74```
75
76## Flywheel Integration
77
78Every `gc ai ask` response includes an implicit feedback mechanism:
79
80```bash
81gc ai ask "how does billing work"
82# Response displayed...
83# Was this helpful? [y/n/e (elaborate)]
84```
85
86Ratings feed directly into the Gluecron AI flywheel, improving answer quality for your specific codebase over time.
87
88## Offline Mode
89
90The CLI caches a compressed semantic index of your repo locally. When offline:
91- `gc ai ask` works against the local index (last synced at most recent `gc push`)
92- Responses are marked `[offline — index from <timestamp>]`
93- Full accuracy resumes on next network connection
94
95## Environment Variables
96
97| Variable | Description |
98|----------|-------------|
99| `GLUECRON_URL` | URL of your Gluecron instance |
100| `GLUECRON_TOKEN` | API token for authentication |
101| `GLUECRON_OFFLINE` | Set to `1` to force offline mode |
102| `GLUECRON_NO_FLYWHEEL` | Set to `1` to disable telemetry (not recommended) |
0103