feat(station): README documents for each packages.#1287
Closed
samchon wants to merge 1675 commits into
Closed
Conversation
* feat(benchmark): formula, explicit type definition. * fix a little bit
…#1067) * feat(format): add Prettier formatting scripts and GitHub Actions workflow for format checks * chore(format): rename workflow from 'build' to 'prettier' * refactor(Logger): rename OutputChannel type for clarity
…1066) * feat(agent): interface schema, also validates DB property existence. * validation feedback logic * fix typo * enhance prompts and fix validation path * path parameters and JWT parameters guiding * fix typo * enhance prompt to reduce time consumption * change config values * fix * fix * enhance prompts * detailed examples and descriptions * fix `stringifyValidateFailure()` * stringifyValidationFailure * bump up agentica * fix missed * another way * try to change constant values * restore config
* feat(agent): change configuration * documentation * format * finish configuration * commit prettierignore * ignore prompt built
* fix(agent): realize agent has to know `MyGlobal`. * fix(agent): realize agent prohibits raw SQL query.
…1097) * feat(biome): add Biome configuration and integrate linting workflow - Introduced a new `biome.json` file for Biome configuration, enabling linting with specified rules. - Updated `package.json` to include Biome as a development dependency and added a lint script. - Modified `pnpm-lock.yaml` to reflect the new Biome dependency and its CLI tools for various platforms. - Created a GitHub Actions workflow for linting TypeScript files using Biome on pull requests. * refactor(biome): clean up configuration by removing unused JavaScript parser settings * refactor(workflow): rename prettier workflow to lint and update job name for clarity
* fix(agent): operation name "index" is reserved. * Update packages/agent/src/orchestrate/interface/programmers/AutoBeInterfaceOperationProgrammer.ts Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * fix prompt and biome * format --------- Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* feat(estimate): add comparison tool with PR review fixes * chore: update pnpm-lock.yaml for estimate package * fix(estimate): address PR review feedback * fix(estimate): resolve biome lint errors * fix(estimate): fix _error reference in base-agent catch block
* feat(estimate): add comparison tool with PR review fixes * chore: update pnpm-lock.yaml for estimate package * fix(estimate): address PR review feedback * fix(estimate): resolve biome lint errors * fix(estimate): fix _error reference in base-agent catch block
* feat(estimate): improve scoring accuracy and agent evaluation' wq * fix: remove unused imports for lint * fix: remove unused imports
* feat(estimate): improve scoring accuracy and agent evaluation' wq * fix: remove unused imports for lint * fix: remove unused imports * docs(estimate): update README with scoring formula and agent changes
* feat(estimate): v0.3.0 scoring improvements and agent enhancements - Document quality: content analysis, README quality, boilerplate detection - Requirements coverage: controller-provider mapping check - Test coverage: describe/it structure detection - Logic completeness: multiline empty method/catch detection - Agent: concurrency limit (max 3), prompt dedup guidelines - Compare reporter: penalty display - CLI: scoreBreakdown in JSON output - Prompts: evaluation guidelines to reduce duplicate issues * fix: complete document-quality evaluator file * fix: replace corrupted document-quality evaluator * chore: bump estimate version to 0.3.0 * feat(estimate): fix type evaluation with Prisma stub generation and gate strictness - TypeEvaluator: read all TS files (includes decorators, utils, guards) - TypeEvaluator: generate Prisma client stub from schema models - Parse model fields with proper TypeScript types - Generate PrismaDelegate, FindArgs, MutateArgs interfaces - Generate per-model utility types (GetPayload, CreateInput, etc.) - context-builder: discover all src/**/*.ts files - Gate: fail on >100 critical/warning type errors or >30% syntax warnings - Gate: exclude suggestions from gate threshold - Gate: display failure reason in terminal - Warning penalty: starts at 30% ratio, max -20 (was 50%, max -10) - Agent: parallel execution, critical cap (20+→60, 10+→70, 5+→80) - Agent: cap display in final score - UI: compact agent output, grouped issues top 5, message truncation * fix: replace any with typed filter in agent critical cap
…1271) * feat(agent): event dispatch from outside of preliminary orchestrate. * fix format * complete aggregation * exact progress tracing * fix format
* feat(agent): preliminary re-write for only critical case. * more concisely
Change all user-visible "Benchmark" labels to "Estimate": - Navigation tab and sidebar label - Dashboard loading/error messages - Heatmap colors aligned to grade boundaries (A/B/C/D/F)
* feat(agent): no more review agents required. * fix website and claude code prompts * fix lint * fix miss edited * enhance compiler feedback hinkts * fix format * concise ts2339 hint * explicit complete type * no re-write required agents * enhance template * fix format * bump up embed-eslint and embed-typescript * fix nullable * fix test failure * fix preliminary controller
* feat(agent): enhance realize template rules. * add validation logic * propagate realize function's template code. * fix logic problem (identifyRootCauseFiles) * fix format * fix template problem
* writing blog article * revised article * complement * writing content * 1st completion * modifying * most completed * complement * 1st completion * about CC * complete in korean * complete article * fix typo and update benchmark.json
* feat(agent): automated agent healing process. * check zero peropties * turn off zero property checking due to gpt-5.4-mini * Update packages/agent/src/orchestrate/interface/utils/AutoBeJsonSchemaValidator.ts Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* feat(agent): 1 retry change for realize functions. * fix format * fix logics * fix prompts * fix title * benchmark summary
* feat(agen): detailed guideline about realize phase. * add test cases * fix format * fix orchestrateRealizeCorrectWithRetry
#1264) * fix(estimate): scoring accuracy fixes, benchmark ESM fix, and error logging - Fix run-benchmark.sh to use ts-node instead of node dist/ (ESM resolution failure on Node.js v24 caused agent evaluation to be silently skipped) - Fix gate multiplier: perfect gate score (100) now yields multiplier 1.0 exactly - Fix penalty rounding: compute effective penalty from exact scaled sum to avoid rounding drift exceeding cap - Extract magic numbers into named constants (GATE_MULTIPLIER_FLOOR, TYPE_CRITICAL_RATIO, PRISMA_PENALTY_CAP, MAX_COMBINED_PENALTY) - Replace safeScenario with safeSuite for golden set crash isolation (preserves partial results) - Add error logging to 4 silent catch blocks in runtime evaluator, golden set evaluator, and markdown reporter - Remove --passWithNoTests from vitest config * chore(estimate): add agent-included benchmark reports + prettier formatting - Update 22 estimate-report.json files with agent evaluation results (a3) - 13 models × 4 projects benchmark with Security/LLMQuality/Hallucination agents - Update dashboard and website benchmark-summary.json - Apply prettier formatting to estimate source files * test(estimate): add unit tests for scoring constants and scenario helpers - Add score.test.ts: 13 tests for PHASE_WEIGHTS, AGENT_WEIGHTS, scoreToGrade, createEmptyPhaseResult, and all scoring constants - Add scenario-helpers.test.ts: 5 tests for pass(), fail(), safeSuite() - Add safeSuite() function for crash-resilient scenario suite execution - Export new scoring constants (GATE_MULTIPLIER_FLOOR, TYPE_CRITICAL_RATIO, PRISMA_PENALTY_CAP, MAX_COMBINED_PENALTY) from types/index.ts * ci: retrigger CI checks * fix: resolve typos, prettier, and lint errors from transformer PR - Fix "initialised" → "initialized" typos in 3 files - Apply prettier formatting to transformer programmer and test files - Suppress biome noExplicitAny for schema union narrowing * refactor(estimate): split pipeline.ts and cli.ts into focused modules - pipeline.ts (902→486 lines): extract penalty logic to penalty.ts, score calculation to score-calculator.ts, reference collection to reference-collector.ts - cli.ts (998→deleted): split into cli/program.ts (command definitions), cli/commands/evaluate.ts, cli/commands/batch.ts, cli/commands/compare.ts, cli/output.ts (print functions), cli/types.ts (interfaces) - All exports preserved via index.ts barrel files - No behavioral changes, build passes * refactor(estimate): split types/score.ts into focused modules - score.ts (345→165 lines): interfaces and types only - constants.ts: all scoring constants, weights, and runtime validations - score-utils.ts: utility functions (scoreToGrade, createEmptyPhaseResult, generateExplanation) - Update index.ts barrel exports and test imports accordingly - No behavioral changes, build passes * test(estimate): add unit tests for penalty and score-calculator modules - penalty.test.ts: 19 tests covering all 5 penalty types, proportional cap scaling, edge cases - score-calculator.test.ts: 31 tests covering weighted scoring, gate multiplier, gate failure partial credit, issue deduplication, grade mapping - vitest.config.ts: package-local config to fix vite-tsconfig-paths ESM resolution error * feat(estimate): add fix advisory system for post-evaluation guidance - Add FixAdvisory types (fix-advice.ts): FixAdvice, CodeSnippet, PenaltyRecovery - Add generateFixAdvisory() with impact estimation and code snippet extraction - Integrate into CLI evaluate command and markdown reporter - Include reference evaluator issues and penalty recovery opportunities - Add 21 unit tests covering all advisory features * feat(estimate): add LLM-powered compile error diagnosis command - Add `diagnose` CLI subcommand for 7-step forensic error analysis - Support both --input (run eval first) and --report (use existing JSON) - Add DIAGNOSE_AGENT.md system prompt based on test/DIAGNOSE.md methodology - Add text mode to LLMClient.chat() for non-JSON responses - Reads Prisma schemas, DTO types, and source code as LLM context - Outputs diagnosis as markdown file (diagnosis.md) * fix(estimate): add Node v22 compatibility for ts-node module resolution Force CJS module resolution via TS_NODE_COMPILER_OPTIONS in run-benchmark.sh to fix Module_not_found errors on Node v22 where extensionless imports in workspace packages fail to resolve. * fix: typo penalised → penalized * fix: remove unused imports (biome lint) * style: apply prettier formatting
* docs(estimate): rewrite README as estimate-specific documentation * feat(estimate): enable agent evaluation by default --use-agent (default false) → --no-agent (default true) Agent evaluation now runs by default. Use --no-agent to disable. * fix: remove accidentally committed benchmark report files
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This pull request adds comprehensive and consistent README documentation for all major packages and applications in the AutoBE monorepo. The new and updated READMEs provide clear overviews, usage instructions, setup guides, and feature summaries, improving onboarding and maintainability for both developers and users.
The most important changes include:
New and Improved Documentation for Applications:
README.mdfiles forhackathon-api,hackathon-server,hackathon-ui,playground-api,playground-server, andplayground-ui, each with project descriptions, setup instructions, environment variables, and usage scripts. [1] [2] [3] [4] [5] [6]New and Improved Documentation for Core Packages:
README.mdfiles for core packages including@autobe/agent,@autobe/benchmark,@autobe/compiler,@autobe/filesystem,@autobe/interface,@autobe/rpc, and@autobe/ui, each describing the package’s role, features, and usage. [1] [2] [3] [4] [5] [6] [7]Major Rewrite and Expansion for
@autobe/estimate:@autobe/estimateREADME, adding a detailed overview, pipeline breakdown, CLI instructions, options tables, and output file descriptions, replacing the previous minimal documentation.These documentation improvements bring consistency and clarity across the codebase, making it much easier for new contributors and users to understand and work with each component.