edsger 0.76.0 → 0.78.0

package/dist/phases/quality-benchmark/types.d.ts

@@ -1,312 +0,0 @@
-/**
- * Type definitions for the quality-benchmark phase.
- *
- * The shapes in this file are the canonical TS counterpart to the JSON
- * structures stored in the `quality_reports` table and emitted by the
- * Claude Agent SDK at the end of a benchmark run. Keep them in sync with:
- *   * supabase/migrations/20260515000000_create_quality_reports.sql
- *   * packages/edsger/src/phases/quality-benchmark/rubric.md
- */
-export declare const DIMENSIONS: readonly ["architecture", "code_quality", "test_coverage", "security", "performance", "documentation", "maintainability", "dependency_health"];
-export type Dimension = (typeof DIMENSIONS)[number];
-export type Grade = 'A' | 'B' | 'C' | 'D' | 'F' | 'N/A';
-export type Severity = 'critical' | 'high' | 'medium' | 'low';
-export type Effort = 'low' | 'medium' | 'high';
-export type Impact = 'low' | 'medium' | 'high';
-export declare const RUBRIC_VERSION: "v1";
-export type Archetype = 'library' | 'cli' | 'web-app' | 'backend-service' | 'mobile' | 'data-pipeline' | 'infra' | 'monorepo' | 'embedded' | 'desktop-app' | 'other';
-export interface DetectedContext {
-    archetype: Archetype;
-    primary_languages: string[];
-    frameworks: string[];
-    package_managers: string[];
-    build_system?: string;
-    test_frameworks: string[];
-    ci_configured: boolean;
-    lockfile_present: boolean;
-    scanned_commit_sha: string;
-    file_count_scanned: number;
-    total_loc_approx: number;
-}
-/** Language tag used to gate tool applicability. `'all'` = polyglot tools. */
-export type LanguageTag = 'js' | 'ts' | 'py' | 'go' | 'rust' | 'java' | 'kotlin' | 'ruby' | 'c' | 'cpp' | 'cs' | 'swift' | 'all';
-export type ToolCategory = 'lint' | 'sast' | 'duplication' | 'complexity' | 'dead-code' | 'cycles' | 'coverage' | 'dep-vuln' | 'dep-outdated' | 'dep-unused' | 'dep-license' | 'secrets' | 'loc-stats' | 'typecheck';
-export type InstallerPrereq = 'pipx' | 'go' | 'cargo' | 'npx' | 'gem' | null;
-export interface ToolCatalogEntry {
-    /** Stable identifier matching `tool_versions` keys and parser names. */
-    id: string;
-    /** Human-readable label for UI / logs. */
-    label: string;
-    category: ToolCategory;
-    /** Languages this tool applies to. */
-    applies_to: readonly LanguageTag[];
-    /** Bash command to detect presence. Should be safe (no side effects). */
-    probe: string;
-    /**
-     * Bash command to install the tool to user-space. null means we never
-     * attempt to install (e.g. project-local-only tools accessed via npx).
-     */
-    install: string | null;
-    /** Prerequisite installer that must itself be present. */
-    install_prereq: InstallerPrereq;
-    /**
-     * Command template. Placeholders:
-     *   %REPO_ROOT%   - absolute path to the repo
-     *   %PKG_MANAGER% - detected package manager (npm/pnpm/yarn)
-     *   %SCAN_DIR%    - per-run scratch dir for tool outputs
-     */
-    command: string;
-    timeout_minutes: number;
-    /** Parser module key under `phases/quality-benchmark/parsers/`. */
-    parser: string;
-    /**
-     * Subscores this tool feeds. A single tool may contribute to multiple
-     * subscores (e.g. semgrep -> security.sast + performance.n_plus_one).
-     */
-    subscores: readonly `${Dimension}.${string}`[];
-    /**
-     * Optional conditional gate that depends on detected_context — used when
-     * applies_to is too coarse (e.g. npm-audit only runs with package-lock.json).
-     */
-    requires?: {
-        file_present?: string[];
-        package_manager?: string[];
-        framework?: string[];
-    };
-    /**
-     * If non-zero exit code is expected even on success (e.g. linters that
-     * exit 1 when findings exist), tolerate it as long as stdout parses.
-     */
-    tolerate_nonzero_exit?: boolean;
-}
-export interface UnavailableTool {
-    name: string;
-    category: ToolCategory;
-    install_command: string | null;
-    reason: 'not_found' | 'wrong_version' | 'install_failed' | 'prereq_missing' | 'install_disabled' | 'timed_out' | 'unknown';
-    /** Tail of stderr when install or probe failed (<=500 chars). */
-    detail?: string;
-}
-export interface ToolRunOutput {
-    tool_id: string;
-    ran_at: string;
-    duration_ms: number;
-    exit_code: number;
-    /** Count of findings parsed from stdout (post-parser). */
-    findings_count: number;
-    /** Short human-readable summary, e.g. "23 errors, 41 warnings". */
-    summary: string;
-    /** True if the tool's parser succeeded; false if output couldn't be parsed. */
-    parsed: boolean;
-    /** Tail of stderr if non-zero exit and not tolerated (<=500 chars). */
-    stderr_tail?: string;
-    /** Path on disk where the full tool output is saved for audit. */
-    raw_output_path?: string;
-}
-/**
- * Tier 1 — counts only. Used for style/quality linters where individual
- * findings are noisy and the score depends on volume, not specific lines.
- * Examples: eslint, ruff, golangci-lint, clippy, rubocop, mypy, tsc.
- */
-export interface ToolSummaryCounts {
-    tier: 'counts';
-    counts: {
-        errors: number;
-        warnings: number;
-        info: number;
-    };
-}
-/**
- * Tier 2 — counts plus the top-N most severe findings with file:line.
- * Used for security and correctness tools where users need to see
- * specific offending lines. Top findings feed the Verifier and the
- * evidence section of the report.
- * Examples: semgrep, gosec, bandit, gitleaks, npm-audit, pip-audit.
- */
-export interface ToolSummaryFindings {
-    tier: 'findings';
-    counts: {
-        critical: number;
-        high: number;
-        medium: number;
-        low: number;
-        total: number;
-    };
-    top_findings: RawFinding[];
-}
-/**
- * Tier 3 — domain-specific structured metrics. Used for tools whose
- * value is statistics or distributions, not findings.
- * Examples: scc (LOC), lizard (complexity), radon (complexity buckets),
- * license-checker (license breakdown), jscpd (duplication %).
- */
-export interface ToolSummaryMetrics {
-    tier: 'metrics';
-    metrics: Record<string, unknown>;
-}
-export type ToolSummary = ToolSummaryCounts | ToolSummaryFindings | ToolSummaryMetrics;
-export interface ParsedToolOutput {
-    tool_id: string;
-    /** Discriminated union of the three summary tiers. */
-    summary: ToolSummary;
-    /** Short, human-readable one-liner, e.g. "23 errors, 41 warnings". */
-    oneliner: string;
-}
-export interface ParserContext {
-    repo_root: string;
-    /** Repo-relative paths in this set are valid (cheap existence check). */
-    files_known?: Set<string>;
-}
-export type ParserFn = (stdout: string, stderr: string, ctx: ParserContext) => ParsedToolOutput;
-export type FindingSource = `tool:${string}` | 'llm_judgment';
-export interface RawFinding {
-    /** Repo-relative path. */
-    file: string;
-    /** 1-based line number. */
-    line: number;
-    /** One-sentence description. */
-    issue: string;
-    severity: Severity;
-    /** Optional code snippet, <=200 chars. */
-    snippet?: string;
-    /** Source attribution. */
-    source: FindingSource;
-    /** Tool-specific rule id when available. */
-    rule_id?: string;
-    /** CWE identifiers when known. */
-    cwe?: string[];
-    /** Subscore this finding feeds into, e.g. "security.sast". */
-    subscore_key: `${Dimension}.${string}`;
-}
-export interface VerifiedFinding extends RawFinding {
-    /** When the same issue is found by multiple tools, all source IDs are merged. */
-    sources: FindingSource[];
-    /** Internal stable id used for cross-referencing in recommendations. */
-    id: string;
-}
-export interface AppliedCheck {
-    /** Stable check id, e.g. "arch.circular_deps". */
-    id: string;
-    /** Tool id from catalog, or "llm_judgment". */
-    tool: string;
-    /** Weight within parent subscore. */
-    weight: number;
-    /** True if the check actually executed and produced parseable output. */
-    measured: boolean;
-    /** Reason it didn't run (when measured = false). */
-    unmeasured_reason?: string;
-}
-export type AppliedChecksByDimension = Record<Dimension, AppliedCheck[]>;
-export interface GitSignals {
-    authors_90d?: number;
-    top_churn_files?: {
-        file: string;
-        commits_30d: number;
-    }[];
-    bug_fix_commits_90d?: number;
-    commit_message_sample?: string[];
-}
-export interface GitHubIssueSignals {
-    open_bugs?: number;
-    open_security?: number;
-    open_regressions?: number;
-    avg_issue_age_days?: number;
-    unavailable?: string;
-}
-export interface SentrySignals {
-    unresolved_7d?: number;
-    top_errors?: {
-        title: string;
-        count: number;
-        first_seen: string;
-        last_seen: string;
-    }[];
-    unavailable?: string;
-}
-export interface ExternalSignals {
-    git?: GitSignals;
-    github_issues?: GitHubIssueSignals;
-    sentry?: SentrySignals;
-}
-export interface Subscore {
-    /** Aggregated 0-100 value for this subscore (or null if N/A). */
-    value: number | null;
-    /** Fraction of weighted checks that actually ran (0..1). */
-    measured_coverage: number;
-    /** Optional human-readable detail, e.g. "Limited measurement…". */
-    summary?: string;
-}
-export interface Recommendation {
-    title: string;
-    effort: Effort;
-    impact: Impact;
-    description: string;
-    files?: string[];
-    /** Optional reference back to the evidence IDs this rec resolves. */
-    blocks_evidence?: string[];
-}
-export interface DimensionScore {
-    score: number | null;
-    grade: Grade;
-    summary: string;
-    subscores: Record<string, Subscore>;
-    evidence: VerifiedFinding[];
-    recommendations: Recommendation[];
-    n_a_reason: string | null;
-}
-export type DimensionScores = Record<Dimension, DimensionScore>;
-export interface QualityReportPayload {
-    rubric_version: typeof RUBRIC_VERSION;
-    detected_context: DetectedContext;
-    tool_versions: Record<string, string>;
-    unavailable_tools: UnavailableTool[];
-    applied_checks: AppliedChecksByDimension;
-    tool_outputs: Record<string, ToolRunOutput>;
-    external_signals: ExternalSignals;
-    dimension_scores: DimensionScores;
-    dropped_findings: number;
-    overall_score: number | null;
-    overall_grade: Grade;
-    executive_summary: string;
-    low_confidence: boolean;
-}
-export type RunStatus = 'pending' | 'running' | 'completed' | 'failed' | 'cancelled';
-export type RunPhase = 'detection' | 'probing' | 'installation' | 'execution' | 'external_signals' | 'verification' | 'synthesis';
-export interface RunProgress {
-    tools_total?: number;
-    tools_completed?: number;
-    current_tool?: string;
-    current_dimension?: Dimension;
-    message?: string;
-}
-export interface QualityReportRow {
-    id: string;
-    product_id: string;
-    commit_sha: string;
-    branch: string | null;
-    repo_root: string | null;
-    rubric_version: string;
-    detected_context: DetectedContext | Record<string, never>;
-    tool_versions: Record<string, string>;
-    unavailable_tools: UnavailableTool[];
-    applied_checks: AppliedChecksByDimension | Record<string, never>;
-    tool_outputs: Record<string, ToolRunOutput>;
-    external_signals: ExternalSignals;
-    dropped_findings: number;
-    dimension_scores: DimensionScores | Record<string, never>;
-    overall_score: number | null;
-    overall_grade: Grade | null;
-    executive_summary: string | null;
-    low_confidence: boolean;
-    status: RunStatus;
-    current_phase: RunPhase | null;
-    progress: RunProgress;
-    error_message: string | null;
-    error_phase: RunPhase | null;
-    started_at: string | null;
-    completed_at: string | null;
-    duration_seconds: number | null;
-    created_by: string | null;
-    created_at: string;
-    updated_at: string;
-}

package/dist/phases/quality-benchmark/types.js

@@ -1,23 +0,0 @@
-/**
- * Type definitions for the quality-benchmark phase.
- *
- * The shapes in this file are the canonical TS counterpart to the JSON
- * structures stored in the `quality_reports` table and emitted by the
- * Claude Agent SDK at the end of a benchmark run. Keep them in sync with:
- *   * supabase/migrations/20260515000000_create_quality_reports.sql
- *   * packages/edsger/src/phases/quality-benchmark/rubric.md
- */
-// ---------------------------------------------------------------------------
-// Rubric: dimensions, grades, severities
-// ---------------------------------------------------------------------------
-export const DIMENSIONS = [
-    'architecture',
-    'code_quality',
-    'test_coverage',
-    'security',
-    'performance',
-    'documentation',
-    'maintainability',
-    'dependency_health',
-];
-export const RUBRIC_VERSION = 'v1';

package/dist/phases/quality-benchmark/workspace.d.ts

@@ -1,31 +0,0 @@
-/**
- * Git workspace prep for `edsger quality-benchmark`.
- *
- * Mirrors the convention in `phases/pr-resolve/workspace.ts`:
- *   - Clones into `~/edsger/quality-<owner>-<repo>`
- *   - Reuses + fetches + hard-resets to the remote default branch when the
- *     dir already has a `.git`
- *   - Authenticates via `buildCredentialArgs(token)` so the GitHub App
- *     installation token never lands in the remote URL or `ps`
- *
- * Quality runs are read-only — no per-commit / per-run subdirectory, one
- * checkout per (owner, repo) is enough and lets repeated runs skip the
- * full clone cost.
- */
-export interface PrepareQualityWorkspaceOptions {
-    owner: string;
-    repo: string;
-    token: string;
-    verbose?: boolean;
-}
-export interface QualityWorkspace {
-    /** Absolute path to the checkout. */
-    repoPath: string;
-    /** The default branch we ended up on (e.g. "main"). */
-    branch: string;
-}
-export declare function getQualityWorkspacePath(owner: string, repo: string): string;
-/**
- * Clone (or reuse) the product's repo and check out its default branch.
- */
-export declare function prepareQualityWorkspace(options: PrepareQualityWorkspaceOptions): QualityWorkspace;

package/dist/phases/quality-benchmark/workspace.js

@@ -1,110 +0,0 @@
-/**
- * Git workspace prep for `edsger quality-benchmark`.
- *
- * Mirrors the convention in `phases/pr-resolve/workspace.ts`:
- *   - Clones into `~/edsger/quality-<owner>-<repo>`
- *   - Reuses + fetches + hard-resets to the remote default branch when the
- *     dir already has a `.git`
- *   - Authenticates via `buildCredentialArgs(token)` so the GitHub App
- *     installation token never lands in the remote URL or `ps`
- *
- * Quality runs are read-only — no per-commit / per-run subdirectory, one
- * checkout per (owner, repo) is enough and lets repeated runs skip the
- * full clone cost.
- */
-import { execFileSync, execSync } from 'child_process';
-import { existsSync } from 'fs';
-import { homedir } from 'os';
-import { join } from 'path';
-import { logError, logInfo, logSuccess } from '../../utils/logger.js';
-import { buildCredentialArgs } from '../pr-resolve/workspace.js';
-export function getQualityWorkspacePath(owner, repo) {
-    return join(homedir(), 'edsger', `quality-${owner}-${repo}`);
-}
-/**
- * Clone (or reuse) the product's repo and check out its default branch.
- */
-export function prepareQualityWorkspace(options) {
-    const { owner, repo, token, verbose } = options;
-    const repoPath = getQualityWorkspacePath(owner, repo);
-    const repoUrl = `https://github.com/${owner}/${repo}.git`;
-    const credArgs = buildCredentialArgs(token);
-    if (existsSync(join(repoPath, '.git'))) {
-        logInfo(`Reusing existing quality workspace at ${repoPath}`);
-        try {
-            execSync(`git remote set-url origin "${repoUrl}"`, {
-                cwd: repoPath,
-                stdio: 'pipe',
-            });
-        }
-        catch {
-            // Non-fatal — the URL may already be correct.
-        }
-        try {
-            execFileSync('git', [...credArgs, 'fetch', 'origin', '--prune'], {
-                cwd: repoPath,
-                stdio: 'pipe',
-            });
-        }
-        catch {
-            logError('Could not fetch latest changes; analysis will use cached state');
-        }
-        const branch = detectDefaultBranch(repoPath, credArgs);
-        try {
-            execFileSync('git', ['checkout', '-B', branch, `origin/${branch}`], {
-                cwd: repoPath,
-                stdio: 'pipe',
-            });
-            execFileSync('git', ['reset', '--hard', `origin/${branch}`], {
-                cwd: repoPath,
-                stdio: 'pipe',
-            });
-        }
-        catch (error) {
-            throw new Error(`Failed to update workspace to origin/${branch}: ${error instanceof Error ? error.message : String(error)}`);
-        }
-        if (verbose) {
-            logInfo(`Workspace synced to origin/${branch}`);
-        }
-        return { repoPath, branch };
-    }
-    logInfo(`Cloning ${owner}/${repo} for quality benchmark...`);
-    try {
-        execFileSync('git', [...credArgs, 'clone', repoUrl, repoPath], {
-            stdio: 'pipe',
-        });
-        logSuccess(`Cloned to ${repoPath}`);
-    }
-    catch (error) {
-        throw new Error(`Failed to clone ${owner}/${repo}: ${error instanceof Error ? error.message : String(error)}`);
-    }
-    const branch = detectDefaultBranch(repoPath, credArgs);
-    return { repoPath, branch };
-}
-/**
- * Return the remote's default branch (the one `origin/HEAD` points at).
- * Falls back to `main` if the symbolic ref isn't set — `git remote set-head`
- * is used to (re)populate it before reading.
- */
-function detectDefaultBranch(repoPath, credArgs) {
-    try {
-        execFileSync('git', [...credArgs, 'remote', 'set-head', 'origin', '-a'], {
-            cwd: repoPath,
-            stdio: 'pipe',
-        });
-    }
-    catch {
-        // Non-fatal; the symbolic ref may already be set from a fresh clone.
-    }
-    try {
-        const out = execFileSync('git', ['symbolic-ref', '--short', 'refs/remotes/origin/HEAD'], { cwd: repoPath, stdio: ['ignore', 'pipe', 'pipe'] }).toString();
-        const ref = out.trim();
-        if (ref.startsWith('origin/')) {
-            return ref.slice('origin/'.length);
-        }
-    }
-    catch {
-        // fall through
-    }
-    return 'main';
-}

package/dist/phases/recipes/index.d.ts

@@ -1,63 +0,0 @@
-/**
- * Recipes phase: clone the product's repo, ask Claude to identify each
- * non-trivial capability the product implements, and persist HOW it's built
- * via the recipes / product_recipes tables.
- *
- * Production-grade behaviours layered on top of the basic agent loop:
- *
- *   - Heartbeat: `last_heartbeat_at` on the recipe_scans row is refreshed
- *     on every assistant message so the reader can detect stalled / crashed
- *     runs (see desktop-app/.../services/db/recipe-scans.ts for the lazy
- *     reaper).
- *   - Cancellation-safe writes: markRunning / markSuccess / markFailed only
- *     touch rows whose status is in {pending, running}. If the user clicked
- *     Stop and the row is now 'cancelled', the final write no-ops.
- *   - Per-call MCP writes: agent commits each create / update / link /
- *     unlink as it goes. There is no "submit at the end" buffer — partial
- *     progress survives even if the agent later errors out.
- */
-import type { SupabaseClient } from '@supabase/supabase-js';
-import type { RecipeSummary } from './types.js';
-export interface RecipesPhaseOptions {
-    /** Product-scoped scan. Mutually exclusive with `repoId`. */
-    productId?: string;
-    /** Repo-only scan: a single repositories row, no product context. */
-    repoId?: string;
-    scanId: string;
-    guidance?: string;
-    verbose?: boolean;
-}
-export interface RecipesPhaseResult {
-    status: 'success' | 'error' | 'cancelled';
-    message: string;
-    counts?: {
-        created: number;
-        updated: number;
-        linked: number;
-        unlinked: number;
-    };
-}
-export declare function runRecipesPhase(options: RecipesPhaseOptions): Promise<RecipesPhaseResult>;
-export declare function getProductTeamId(supabase: SupabaseClient, productId: string): Promise<string | null>;
-export declare function getScanCreator(supabase: SupabaseClient, scanId: string): Promise<{
-    created_by: string;
-} | null>;
-export declare function listTeamRecipes(supabase: SupabaseClient, teamId: string): Promise<RecipeSummary[]>;
-export declare function listProductRecipeLinks(supabase: SupabaseClient, productId: string): Promise<{
-    recipe_id: string;
-    name: string;
-}[]>;
-export declare function listRepositoryRecipeLinks(supabase: SupabaseClient, repositoryId: string): Promise<{
-    recipe_id: string;
-    name: string;
-}[]>;
-/**
- * Claim the row by flipping `pending` → `running`. Returns true on success
- * (we won the claim) and false when the row has already moved on (e.g. user
- * cancelled before the CLI started). Bounded by the status filter so we
- * can't accidentally resurrect a 'cancelled' row.
- */
-export declare function markRunning(supabase: SupabaseClient, scanId: string): Promise<boolean>;
-export declare function heartbeat(supabase: SupabaseClient, scanId: string): Promise<void>;
-export declare function markFailed(supabase: SupabaseClient, scanId: string, errorMessage: string): Promise<boolean>;
-export declare function markSuccess(supabase: SupabaseClient, scanId: string): Promise<boolean>;