Job application tracker and AI-powered job search assistant. Helps job seekers manage their search journey with AI resume review, job matching, task logging, and application analytics, all while keeping your data private.
TypeScript ai-agentai-resume-analyzerapplication-trackerapplication-tracking-systematscareerjob-applicationjob-automationjob-huntingjob-searchjob-seeker-toolsjob-trackerjob-trackingjobsearchjobseekersmcpollamaresumeresume-reviewself-hosted
import { z } from "zod";
import prisma from "@/lib/db";
import { APP_CONSTANTS } from "@/lib/constants";
import { McpSaveMatchResultSchema } from "@/models/mcp.schema";
import { checkMcpRateLimit } from "@/lib/mcp/rate-limit";
import { parseJobMatch } from "@/lib/ai/jobMatch/parse";
import type { JobMatchData } from "@/models/ai.schemas";
import type { DescriptionCompleteness } from "@/models/job.model";
export async function handleSaveMatchResult(
input: z.infer<typeof McpSaveMatchResultSchema>,
userId: string,
tokenName: string,
): Promise<{ content: Array<{ type: "text"; text: string }> }> {
// Shares add_job's bucket by design. At the exact window boundary a match can
// be computed but rejected here; accepted — the user can retry or match
// in-app, and nothing is left in a corrupt state.
const rateCheck = checkMcpRateLimit(userId);
if (!rateCheck.allowed) {
const resetSec = Math.ceil(rateCheck.resetIn / 1000);
return {
content: [{ type: "text", text: `Rate limit exceeded. Try again in ${resetSec}s.` }],
};
}
const parsed = parseJobMatch(input.matchText);
if (!parsed.scores) {
return {
content: [
{
type: "text",
text:
"Could not parse a SCORES line. Include a leading 'SCORES: " +
"match=<0-100> recommendation=<strong|good|partial|weak>' line.",
},
],
};
}
// Provenance needs only id + title, so avoid the full resume graph that
// getDefaultResumeForUser loads for preprocessing. Prefer the resumeId
// add_job handed the agent (the resume actually scored against); the
// profile.userId scope bounds it to the user's own resumes. Fall back to the
// current default if the agent didn't echo an id or it no longer resolves.
let resume =
input.resumeId != null
? await prisma.resume.findFirst({
where: { id: input.resumeId, profile: { userId } },
select: { id: true, title: true },
})
: null;
if (!resume) {
const user = await prisma.user.findUnique({
where: { id: userId },
select: { defaultResumeId: true },
});
resume = user?.defaultResumeId
? await prisma.resume.findFirst({
where: { id: user.defaultResumeId, profile: { userId } },
select: { id: true, title: true },
})
: null;
}
// Same scope as the update below, so a job the caller can't write to
// never leaks its completeness through this read either.
const job = await prisma.job.findFirst({
where: { id: input.jobId, userId, createdVia: { not: null } },
select: { descriptionCompleteness: true },
});
const matchData: JobMatchData = {
matchScore: parsed.scores.matchScore,
recommendation: parsed.scores.recommendation,
body: parsed.body,
resumeId: resume?.id,
resumeTitle: resume?.title,
matchedAt: new Date().toISOString(),
provider: APP_CONSTANTS.MCP_MATCH_PROVIDER_MARKER,
model: tokenName,
analyzed: true,
descriptionCompleteness:
(job?.descriptionCompleteness as DescriptionCompleteness | null) ?? undefined,
};
try {
await prisma.job.update({
where: { id: input.jobId, userId, createdVia: { not: null } },
data: {
matchScore: parsed.scores.matchScore,
matchData: JSON.stringify(matchData),
},
});
} catch (error: any) {
if (error?.code === "P2025") {
return {
content: [
{
type: "text",
text: "Job not found, not owned by this token's user, or not eligible for a match via MCP.",
},
],
};
}
return {
content: [{ type: "text", text: `Error: ${error?.message ?? "Unknown error"}` }],
};
}
return {
content: [
{
type: "text",
text: `Match saved for job ${input.jobId}: ${parsed.scores.recommendation} (score ${parsed.scores.matchScore}).`,
},
],
};
}