[PROD RELEASE] - AI Review/Updates & Fixes by kkartunov · Pull Request #7 · topcoder-platform/tc-ai-api

kkartunov · 2026-03-25T06:08:06Z

No description provided.

PM-4214 - jd workflow init

optimized review context extraction -> dev

wipro provider -> dev

Copilot

Pull request overview

This PR expands the AI skill set documentation and adds new JD rewriting skills while refactoring challenge-context extraction into decomposed, validated calls with structured-output fallbacks. It also introduces a Wipro provider integration, adds integration testing scaffolding, and updates runtime/test configuration.

Changes:

Add/extend SKILL documentation for tech stack, submission guidelines, codebase detection, and new JD rewriting/extraction/formatting skills.
Refactor challenge-context extraction into 4 focused calls + post-validation helpers; add a generic structured-output fallback wrapper.
Add Vitest integration config + integration tests/fixtures; switch agents/scorers to Wipro provider and update build/runtime configs.

Reviewed changes

Copilot reviewed 25 out of 28 changed files in this pull request and generated 13 comments.

Show a summary per file

File	Description
workspace/skills/tech-stack-extraction/SKILL.md	Clarifies exclusions for tech-stack extraction (platform tooling, generic categories, conditional tech).
workspace/skills/submission-guidelines-extraction/SKILL.md	Tightens whatToSubmit rules; adds strong consistency rules for isPatchOfExisting and common mistakes tables.
workspace/skills/jd-structure-formatting/SKILL.md	New skill spec defining canonical JD section order and Markdown formatting rules.
workspace/skills/jd-skills-extraction/SKILL.md	New skill spec for extracting canonical-cased skill keywords from JDs.
workspace/skills/jd-content-rewriting/references/SIGNAL-WORDS.md	Adds reference list of signal words for requirement vs nice-to-have classification.
workspace/skills/jd-content-rewriting/SKILL.md	New skill spec for rewriting raw JDs into professional, specific language.
workspace/skills/codebase-detection/SKILL.md	Refines greenfield vs existing artifact heuristics and adds clearer examples.
vitest.integration.config.ts	Adds a dedicated Vitest config for integration tests.
tsconfig.json	Includes tests in TS project include globs.
tests/integration/challenge-context/extraction.test.ts	Adds integration tests for AI extraction + deterministic validation helpers.
tests/fixtures/github-skills-import.json	Adds a fixture challenge spec + expected context assertions.
src/utils/structured-output-wrapper.ts	Adds multi-strategy structured-output generation fallback + token usage metrics helpers.
src/utils/providers/wipro.ts	Adds Wipro AI SDK provider initialization and default JSON mode settings.
src/utils/index.ts	Re-exports Wipro provider and adjusts exports ordering.
src/utils/auth/m2m.service.ts	Switches m2m import to createRequire-based loading.
src/mastra/workspaces/ai.workspace.ts	Adjusts workspace index paths configuration.
src/mastra/workflows/jd/jd-autowrite-workflow.ts	Adds a JD autowrite workflow using structured output fallback.
src/mastra/workflows/challenge/challenge-context-workflow.ts	Decomposes extraction into focused calls; adds buildChallengeText + validation helpers; changes scorecard base env var.
src/mastra/scorers/skills-matching-scorers.ts	Renames exported scorers/aggregate object (skillDiscovery* → instance*).
src/mastra/index.ts	Registers new JD workflow/agent; switches scorer imports; adds server host env; configures bundler externals/transpile.
src/mastra/agents/skills/skills-matching-agent.ts	Switches model provider to Wipro; switches scorer imports and eval sample env var.
src/mastra/agents/jd/jd-rewriter-agent.ts	Adds JD rewriter agent using Wipro model.
src/mastra/agents/challenge/challenge-parser-agent.ts	Switches challenge parser model provider from Ollama to Wipro.
package.json	Adds integration-test script; bumps Mastra/AI/Vitest deps; adds Wipro provider dependency (git URL).
Dockerfile	Installs git in runtime image.
.nvmrc	Bumps Node version.
.gitignore	Adds test-data to ignored paths.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-25T07:48:00Z

-} from './scorers/skills-matching-scorers';
+  instanceAnswerRelevancyScorer,
+  instancePromptAlignmentScorer,
+} from './scorers/instance-scorers';


The imports reference ./scorers/instance-scorers, but the diff only shows changes to src/mastra/scorers/skills-matching-scorers.ts (and no new instance-scorers module). This will fail at build/runtime. Fix by either (a) adding an instance-scorers.ts file that exports these symbols, or (b) updating imports (here and in skills-matching-agent.ts) to import from the actual scorer module file.

Suggested change

} from './scorers/instance-scorers';

} from './scorers/skills-matching-scorers';

Copilot · 2026-03-25T07:48:01Z

 // ---------------------------------------------------------------------------

-const SCORECARD_API_BASE = process.env.TC_API_BASE_URL ?? 'https://api.topcoder.com/v6';
+const SCORECARD_API_BASE = process.env.TC_API_BASE ?? 'https://api.topcoder.com/v6';


This changes the env var from TC_API_BASE_URL to TC_API_BASE, which is a breaking operational change (existing deployments may only set the former). To avoid production misconfiguration, support both (prefer the new name but fall back to the old one) and/or document the migration clearly.

Suggested change

const SCORECARD_API_BASE = process.env.TC_API_BASE ?? 'https://api.topcoder.com/v6';

const SCORECARD_API_BASE =

process.env.TC_API_BASE ?? process.env.TC_API_BASE_URL ?? 'https://api.topcoder.com/v6';

Copilot · 2026-03-25T07:48:01Z

+import { createRequire } from 'node:module';
 import { M2mConfig } from '../../config/m2m.config';

+const require = createRequire(import.meta.url);
+const m2mAuth = require('tc-core-library-js').auth.m2m;


createRequire(import.meta.url) only works in ESM. If this project is built/loaded as CommonJS (package.json does not show type: module in the diff), import.meta can be undefined and this will crash at runtime. Consider a dual-mode approach (use global require when available; otherwise use createRequire(import.meta.url)), or keep the prior require(...) approach if the runtime is CommonJS.

Suggested change

import { createRequire } from 'node:module';

import { M2mConfig } from '../../config/m2m.config';

const require = createRequire(import.meta.url);

const m2mAuth = require('tc-core-library-js').auth.m2m;

import tcCoreLibrary from 'tc-core-library-js';

import { M2mConfig } from '../../config/m2m.config';

const m2mAuth = tcCoreLibrary.auth.m2m;

Copilot · 2026-03-25T07:48:01Z

+const TECH_STACK_EXCLUDE_PATTERNS: RegExp[] = [
+    // Review / CI / quality-gate tooling
+    /\bsast\b/i,
+    /\bvulnerability\s*scanner/i,
+    /\bcode\s*quality\s*gate/i,
+    /\bstatic\s*analysis/i,
+    /\breview\s*(bot|tool|workflow|system)/i,
+    // Topcoder platform internals
+    /\bcopilot\b/i,
+    /\bscorecard\b/i,
+    /\btopcoder\b/i,
+    // Generic non-technology categories
+    /^security$/i,
+    /^testing$/i,
+    /^ci\/cd$/i,
+    /^documentation$/i,
+];


The pattern /\\btopcoder\\b/i is very broad and will remove any tech_stack item containing 'Topcoder' (e.g., 'Topcoder API', 'Topcoder Skills API'), which can be a real integration technology required by the submission. Consider narrowing this to explicit platform tooling terms (e.g., 'Topcoder Review Bot', 'Topcoder submission portal') rather than the vendor name itself, to avoid false negatives.

Copilot · 2026-03-25T07:48:02Z

+    const raw = readFileSync(
+        resolve(__dirname, '../../fixtures', `${name}.json`),
+        'utf-8',
+    );


__dirname is not defined in ESM, and Vitest commonly runs tests in an ESM context depending on configuration. This can cause the integration tests to fail to load fixtures in some environments. Use an import.meta.url-based directory resolution (e.g., via fileURLToPath) to make fixture loading reliable across module modes.

Copilot · 2026-03-25T07:48:03Z

+    autoIndexPaths: [''],
+    skills: ['skills'],


Using an empty string as a path is ambiguous and relies on implicit path resolution rules in the Workspace implementation. For clarity and to reduce edge cases across OS/path-join behavior, use an explicit current-directory marker (e.g., '.') and an explicit relative skills path (e.g., './skills') if the API expects relative paths.

Suggested change

autoIndexPaths: [''],

skills: ['skills'],

autoIndexPaths: ['.'],

skills: ['./skills'],

Copilot · 2026-03-25T07:48:03Z

+    strategy: StructuredOutputStrategy;
+}
+
+export function toNonNegativeInt(value: unknown): number | null {


This file introduces multiple non-trivial behaviors (token parsing/estimation, compatibility-error classification, and a multi-strategy structured-output fallback loop), but no unit tests are included in this PR. Add focused tests for toNonNegativeInt, buildTokenUsageMetrics (native vs mixed vs estimated), isStructuredOutputCompatibilityError, and the strategy ordering/behavior in generateWithStructuredOutputFallback to prevent regressions.

Copilot · 2026-03-25T07:48:03Z

+    };
+}
+
+export function buildTokenUsageMetrics(


This file introduces multiple non-trivial behaviors (token parsing/estimation, compatibility-error classification, and a multi-strategy structured-output fallback loop), but no unit tests are included in this PR. Add focused tests for toNonNegativeInt, buildTokenUsageMetrics (native vs mixed vs estimated), isStructuredOutputCompatibilityError, and the strategy ordering/behavior in generateWithStructuredOutputFallback to prevent regressions.

Copilot · 2026-03-25T07:48:03Z

+    return modelId.includes('gemini') || modelId.includes('mojo');
+}
+
+export function isStructuredOutputCompatibilityError(error: unknown): boolean {


This file introduces multiple non-trivial behaviors (token parsing/estimation, compatibility-error classification, and a multi-strategy structured-output fallback loop), but no unit tests are included in this PR. Add focused tests for toNonNegativeInt, buildTokenUsageMetrics (native vs mixed vs estimated), isStructuredOutputCompatibilityError, and the strategy ordering/behavior in generateWithStructuredOutputFallback to prevent regressions.

Copilot · 2026-03-25T07:48:03Z

+    );
+}
+
+export async function generateWithStructuredOutputFallback<Schema extends z.ZodTypeAny>({


This file introduces multiple non-trivial behaviors (token parsing/estimation, compatibility-error classification, and a multi-strategy structured-output fallback loop), but no unit tests are included in this PR. Add focused tests for toNonNegativeInt, buildTokenUsageMetrics (native vs mixed vs estimated), isStructuredOutputCompatibilityError, and the strategy ordering/behavior in generateWithStructuredOutputFallback to prevent regressions.

kkartunov and others added 12 commits March 11, 2026 11:43

jd workflow init

901b560

Merge pull request #3 from topcoder-platform/jd-workflow

95dc2c9

PM-4214 - jd workflow init

optimized review context extraction

e1a9c2d

Merge pull request #4 from topcoder-platform/better-context-extration

830ac26

optimized review context extraction -> dev

optimized review context extraction

55368fa

wipro provider

866c597

Merge pull request #5 from topcoder-platform/wipro-provider

ab5be37

wipro provider -> dev

fix git usage for deps

fb9f46e

fix required ESM

dad6801

contradiction at review context generation

2671eff

npm update & switch to wipro provider

f3546bb

try bind to HOST

865fed1

kkartunov requested a review from Copilot March 25, 2026 07:38

Copilot AI reviewed Mar 25, 2026

View reviewed changes

kkartunov merged commit 94236f2 into master Mar 26, 2026
4 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[PROD RELEASE] - AI Review/Updates & Fixes#7

[PROD RELEASE] - AI Review/Updates & Fixes#7
kkartunov merged 12 commits intomasterfrom
develop

kkartunov commented Mar 25, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	} from './scorers/instance-scorers';
	} from './scorers/skills-matching-scorers';

	const SCORECARD_API_BASE = process.env.TC_API_BASE ?? 'https://api.topcoder.com/v6';
	const SCORECARD_API_BASE =
	process.env.TC_API_BASE ?? process.env.TC_API_BASE_URL ?? 'https://api.topcoder.com/v6';

Conversation

kkartunov commented Mar 25, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants