[DEV-1438] Wire CLI to session-scoped judge-facts endpoints

[alexeyzimarev]

Summary

Matches the server-side pivot in kurrent-io/Kurrent.Capacitor#476: judge facts are now scoped to the session's project (repo) and shared among all team members working on it, rather than siloed per-user.

Changes

• Endpoint URLs:
  • `GET /api/judge-facts?category=X` → `GET /api/sessions/{id}/judge-facts?category=X`
  • `POST /api/judge-facts` → `POST /api/sessions/{id}/judge-facts`
• `JudgeFactPayload` drops `source_session_id` (URL-bound now).
• Prompt template rewritten to frame retained facts as project-level observations, with updated examples and explicit anti-examples:
  • ✅ "This repo's tests rely on Testcontainers, so missing Docker is a frequent failure mode"
  • ✅ "This codebase prefers handler-per-file over mega-handlers"
  • ❌ "Alice tends to force-push" (individual — no longer appropriate under repo scoping)
• `FetchAllJudgeFactsAsync` takes `sessionId` so per-category GETs hit the correct session-scoped URL.

Sessions without a detected repo get empty fact lists from the server, so judges silently see "(no patterns retained yet)" — no special-casing on the CLI.

Depends on kurrent-io/Kurrent.Capacitor#476.

Test plan

• `dotnet build src/kapacitor/kapacitor.csproj` — clean, 0 warnings
• `dotnet publish -c Release` — zero IL3050/IL2026 warnings (AOT-clean)
• Full unit suite — 205/205 pass
  • Existing `ParseVerdict`, `ExtractRetainFact`, `Aggregate`, `FormatKnownPatterns`, `BuildQuestionPrompt` tests are server-independent and didn't need changes
• CI
• End-to-end once #476 lands

🤖 Generated with Claude Code

[linear]

DEV-1438 Pivot judge-facts scoping from per-user to per-repo

[qodo-code-review]

Review Summary by Qodo

Wire CLI to session-scoped judge-facts endpoints

✨ Enhancement

Walkthroughs

Description

• Migrate judge-facts endpoints from user-scoped to session-scoped (repo-level)
• Update API URLs to include sessionId in endpoint paths
• Remove SourceSessionId from JudgeFactPayload (now URL-bound)
• Rewrite prompt template to emphasize project-level patterns over user-level
• Update examples to reflect codebase conventions and anti-examples for individual behavior

Diagram

flowchart LR
  A["User-scoped<br/>judge-facts"] -->|"Migrate to"| B["Session-scoped<br/>judge-facts"]
  B -->|"URL includes"| C["sessionId parameter"]
  D["JudgeFactPayload<br/>with SourceSessionId"] -->|"Remove field"| E["Payload without<br/>SourceSessionId"]
  F["User/individual<br/>patterns"] -->|"Reframe as"| G["Project/codebase<br/>patterns"]

Loading

File Changes

1. src/kapacitor/Commands/EvalCommand.cs ✨ Enhancement +11/-10

Wire session-scoped judge-facts API endpoints

• Updated FetchAllJudgeFactsAsync to accept sessionId parameter and construct session-scoped
 endpoint URLs
• Modified PostJudgeFactAsync to use session-scoped endpoint and removed SourceSessionId from
 payload
• Updated comments to reflect DEV-1438 and explain session-scoped repo-level fact storage
• Changed method call to pass sessionId to both fetch and post operations

src/kapacitor/Commands/EvalCommand.cs

---

2. src/kapacitor/Models.cs ✨ Enhancement +6/-8

Remove SourceSessionId from JudgeFactPayload

• Removed SourceSessionId property from JudgeFactPayload record
• Updated XML documentation to explain session-scoped repo-level fact storage
• Clarified that facts are derived from session repo scope and shared across team members

src/kapacitor/Models.cs

---

3. src/kapacitor/Resources/prompt-eval-question.txt 📝 Documentation +8/-7

Reframe judge-facts prompt guidance to project-level scope

• Changed section header from "Known patterns" to "Known patterns for this project"
• Rewrote description to emphasize facts are project/codebase-level, not user-level
• Updated positive examples to focus on codebase conventions (Testcontainers, handler-per-file
 patterns, env vars)
• Added explicit anti-example showing individual behavior ("Alice tends to force-push") is no longer
 appropriate
• Clarified that facts are shared across all evaluators on the same repository

src/kapacitor/Resources/prompt-eval-question.txt

---

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

1. Unescaped sessionId in GET ☑ 🐞 ☼

Description

FetchAllJudgeFactsAsync interpolates the raw sessionId into the URL path for GET
/api/sessions/{id}/judge-facts, so session IDs containing reserved URL characters (e.g. '/') will
produce an invalid path and fail to load known patterns.

Code

src/kapacitor/Commands/EvalCommand.cs[254]

+                using var resp = await httpClient.GetWithRetryAsync($"{baseUrl}/api/sessions/{sessionId}/judge-facts?category={category}");

Evidence

The eval command takes a positional sessionId token verbatim (no escaping/validation) and uses it
directly inside the new session-scoped judge-facts path. If the token contains reserved path
characters, it will be interpreted as additional path segments by the HTTP client/server, breaking
fact retrieval for that sessionId.

src/kapacitor/Commands/EvalCommand.cs[249-256]
src/kapacitor/ArgParsing.cs[10-29]
src/kapacitor/Program.cs[120-137]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`FetchAllJudgeFactsAsync` builds the judge-facts URL via string interpolation: `/api/sessions/{sessionId}/judge-facts...`. Because `sessionId` comes from CLI args/env without validation, reserved URL characters (notably `/`) can corrupt the request path and cause judge-facts retrieval to fail.

## Issue Context
This PR introduced new session-scoped judge-facts endpoints, which adds new places where `sessionId` is embedded in the URL path.

## Fix Focus Areas
- src/kapacitor/Commands/EvalCommand.cs[249-268]

## Suggested fix
- Encode `sessionId` as a path segment, e.g.:
 - `var sid = Uri.EscapeDataString(sessionId);`
 - use `$"{baseUrl}/api/sessions/{sid}/judge-facts?category={Uri.EscapeDataString(category)}"`
- (Optional) centralize URL construction in a helper to avoid repeating escaping logic.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

ⓘ The new review experience is currently in Beta. Learn more