feat(job-search): extract keywords from LLM match call instead of heuristics

Piggybacks keyword extraction onto the existing CV-to-job LLM call —
no extra API calls. The system prompt now instructs the model to return
8-12 English job-search terms (job titles, technologies, skills, domains)
in a new `keywords` field alongside the existing score/summary fields.

Keywords flow: LLM JSON → JobMatchResponse.Keywords → CreateJobSearchTokenRequest →
JobSearchTokenEntity.Keywords (stored comma-separated) → JobSearchSessionEntity.Keywords
(copied at session-creation time, no RAG call needed).

Changes:
- Add Keywords to JobMatchResponse, CreateJobSearchTokenRequest, JobSearchTokenEntity
- IJobTokenService.CreateTokenAsync now accepts IReadOnlyList<string> keywords
- JobTokenService: store keywords on token; TriggerStartAsync reads token.Keywords
  instead of fetching CV text from RAG — removes IRagApiClient dependency
- Remove heuristic ExtractKeywords method
- Migration AddKeywordsToJobSearchTokens: adds Keywords column to cvSearch.JobSearchTokens
- Migration UpdateCvMatchSystemPromptKeywords: updates ai.cv-match.system-prompt seed
  to include keywords in the JSON shape

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-05-29 12:44:13 +03:00
parent a467fac35d
commit b78ede23cf
13 changed files with 462 additions and 39 deletions
@@ -0,0 +1,49 @@
using Microsoft.EntityFrameworkCore.Migrations;
using CvMatcher.Data;
#nullable disable
namespace CvMatcher.Data.Migrations
{
/// <inheritdoc />
public partial class UpdateCvMatchSystemPromptKeywords : Migration
{
private const string OldPrompt =
"You are a strict CV-to-job matching engine. Return JSON only. Score realistically from 0 to 100.\n" +
"Penalize missing required skills. Do not invent experience. Use concise business language.\n" +
"Respond entirely in {{languageName}} — all text fields in the JSON must be in {{languageName}}.\n" +
"JSON shape: {\"score\":number,\"summary\":\"...\",\"strengths\":[\"...\"],\"gaps\":[\"...\"],\"recommendations\":[\"...\"],\"evidence\":[\"...\"]}";
private const string NewPrompt =
"You are a strict CV-to-job matching engine. Return JSON only. Score realistically from 0 to 100.\n" +
"Penalize missing required skills. Do not invent experience. Use concise business language.\n" +
"Respond entirely in {{languageName}} — all text fields in the JSON must be in {{languageName}}.\n" +
"Also extract 8 to 12 English job search keywords from the CV — job titles, technologies, skills, and domains.\n" +
"The keywords array must always be in English regardless of {{languageName}}. Exclude names, emails, phone numbers, and locations.\n" +
"JSON shape: {\"score\":number,\"summary\":\"...\",\"strengths\":[\"...\"],\"gaps\":[\"...\"],\"recommendations\":[\"...\"],\"evidence\":[\"...\"],\"keywords\":[\"term1\",\"term2\"]}";
/// <inheritdoc />
protected override void Up(MigrationBuilder migrationBuilder)
{
migrationBuilder.UpdateData(
schema: MigrationConstants.SchemaName,
table: "AiPrompts",
keyColumns: ["Key", "Language"],
keyValues: new object[] { "ai.cv-match.system-prompt", "*" },
column: "Value",
value: NewPrompt);
}
/// <inheritdoc />
protected override void Down(MigrationBuilder migrationBuilder)
{
migrationBuilder.UpdateData(
schema: MigrationConstants.SchemaName,
table: "AiPrompts",
keyColumns: ["Key", "Language"],
keyValues: new object[] { "ai.cv-match.system-prompt", "*" },
column: "Value",
value: OldPrompt);
}
}
}