# API — Project Search (AI)

> **Base URL:** `https://console.gisportal.id/api/v1`
> **Auth:** `Authorization: Bearer <token>` + scope `project-search:read`
> **Status:** Fase 4 — endpoint terdaftar, implementasi belum aktif

---

## GET /project-search/sources

Daftar sumber data yang tersedia untuk pencarian.

**Response 200:**
```json
{
    "data": [
        {"id": "project", "name": "Project Files", "path": "\\\\10.100.1.10\\project", "status": "not_indexed"},
        {"id": "marketing", "name": "Marketing Files", "status": "not_indexed"}
    ]
}
```

---

## GET /project-search/index/status

Status indeks per source.

**Response 200:**
```json
{
    "data": [
        {"source": "project", "total_files": 0, "indexed_files": 0, "last_indexed_at": null}
    ]
}
```

---

## POST /project-search/ask

Query natural language ke project files.

**Request:**
```json
{
    "query": "dokumen IGN untuk proyek Kendal",
    "sources": ["project"],
    "limit": 5
}
```

**Response 200 (saat aktif):**
```json
{
    "answer": "Dokumen IGN untuk proyek Kendal tersedia di...",
    "sources": [
        {"file": "\\\\10.100.1.10\\project\\2025\\Kendal\\IGN.pdf", "score": 0.95}
    ]
}
```

---

## Catatan

Implementasi lengkap memerlukan:
1. Indexer worker di VM 10.100.1.26 (crawl + embed)
2. pgvector di `geonet_project_search` PostgreSQL
3. Ollama model `nomic-embed-text` untuk embedding
4. Ollama model `qwen2.5` untuk answer generation

Lihat `server/ollama-nas-project-search-requirements.md` untuk requirement detail.