Managing Qdrant

---
name: managing-qdrant
description: |
  Use when working with Qdrant — qdrant vector database management, collection
  health monitoring, shard distribution analysis, and query optimization. Covers
  collection configuration, HNSW index parameters, quantization settings,
  snapshot management, and cluster consensus state. Read this skill before any
  Qdrant operations.
connection_type: qdrant
preload: false
---

# Qdrant Management Skill

Monitor, analyze, and optimize Qdrant vector database instances safely.

## MANDATORY: Discovery-First Pattern

**Always check cluster health and list collections before any query or indexing operations. Never assume collection names, vector dimensions, or payload schemas.**

### Phase 1: Discovery

```bash
#!/bin/bash

QDRANT_URL="${QDRANT_URL:-http://localhost:6333}"
QDRANT_AUTH="${QDRANT_API_KEY:+api-key: $QDRANT_API_KEY}"

qd_get() {
    curl -s ${QDRANT_AUTH:+-H "$QDRANT_AUTH"} "$QDRANT_URL$1"
}

echo "=== Health Check ==="
qd_get "/healthz"

echo ""
echo "=== Telemetry / Version ==="
qd_get "/telemetry" | python3 -c "
import sys, json
data = json.load(sys.stdin)
app = data.get('result', {}).get('app', {})
print(f\"Version: {app.get('version','?')}\")
print(f\"Startup: {app.get('startup','?')}\")
collections = data.get('result', {}).get('collections', {})
print(f\"Total collections: {collections.get('number_of_collections','?')}\")
" 2>/dev/null

echo ""
echo "=== Collections ==="
qd_get "/collections" | python3 -c "
import sys, json
data = json.load(sys.stdin)
for c in data.get('result', {}).get('collections', []):
    print(f\"Collection: {c['name']}\")
" 2>/dev/null

echo ""
echo "=== Collection Details ==="
for coll in $(qd_get "/collections" | python3 -c "
import sys, json
for c in json.load(sys.stdin).get('result',{}).get('collections',[]):
    print(c['name'])
" 2>/dev/null); do
    qd_get "/collections/$coll" | python3 -c "
import sys, json
r = json.load(sys.stdin).get('result', {})
config = r.get('config', {})
params = config.get('params', {})
vectors = params.get('vectors', {})
if isinstance(vectors, dict) and 'size' in vectors:
    print(f\"  {r.get('name','?')}: dim={vectors['size']} distance={vectors.get('distance','?')} points={r.get('points_count',0)} status={r.get('status','?')}\")
else:
    print(f\"  Collection: points={r.get('points_count',0)} status={r.get('status','?')}\")
    for vname, vconf in (vectors if isinstance(vectors, dict) else {}).items():
        print(f\"    vector '{vname}': dim={vconf.get('size','?')} distance={vconf.get('distance','?')}\")
" 2>/dev/null
done

echo ""
echo "=== Cluster Info ==="
qd_get "/cluster" | python3 -c "
import sys, json
data = json.load(sys.stdin).get('result', {})
print(f\"Status: {data.get('status','?')}\")
print(f\"Peer ID: {data.get('peer_id','?')}\")
print(f\"Peers: {len(data.get('peers', {}))}\")
" 2>/dev/null
```

**Phase 1 outputs:** Version, collection list with dimensions and point counts, cluster status.

### Phase 2: Analysis

```bash
#!/bin/bash

QDRANT_URL="${QDRANT_URL:-http://localhost:6333}"
QDRANT_AUTH="${QDRANT_API_KEY:+api-key: $QDRANT_API_KEY}"
COLLECTION="${1:-my_collection}"

qd_get() {
    curl -s ${QDRANT_AUTH:+-H "$QDRANT_AUTH"} "$QDRANT_URL$1"
}

echo "=== Collection Config: $COLLECTION ==="
qd_get "/collections/$COLLECTION" | python3 -c "
import sys, json
r = json.load(sys.stdin).get('result', {})
config = r.get('config', {})
hnsw = config.get('hnsw_config', {})
quant = config.get('quantization_config')
opt = config.get('optimizer_config', {})
print(f\"Points: {r.get('points_count',0)} | Indexed: {r.get('indexed_vectors_count',0)} | Segments: {r.get('segments_count',0)}\")
print(f\"HNSW: m={hnsw.get('m','?')} ef_construct={hnsw.get('ef_construct','?')} full_scan_threshold={hnsw.get('full_scan_threshold','?')}\")
print(f\"Quantization: {quant if quant else 'none'}\")
print(f\"Optimizer: indexing_threshold={opt.get('indexing_threshold','?')} memmap_threshold={opt.get('memmap_threshold','?')}\")
" 2>/dev/null

echo ""
echo "=== Shard Distribution ==="
qd_get "/collections/$COLLECTION/cluster" | python3 -c "
import sys, json
data = json.load(sys.stdin).get('result', {})
local = data.get('local_shards', [])
remote = data.get('remote_shards', [])
print(f\"Local shards: {len(local)} | Remote shards: {len(remote)}\")
for s in local:
    print(f\"  Shard {s.get('shard_id','?')}: points={s.get('points_count',0)} state={s.get('state','?')}\")
" 2>/dev/null

echo ""
echo "=== Payload Indexes ==="
qd_get "/collections/$COLLECTION" | python3 -c "
import sys, json
r = json.load(sys.stdin).get('result', {})
pi = r.get('payload_schema', {})
for field, info in pi.items():
    print(f\"  {field}: type={info.get('data_type','?')} indexed={info.get('points',0)} points\")
" 2>/dev/null

echo ""
echo "=== Snapshots ==="
qd_get "/collections/$COLLECTION/snapshots" | python3 -c "
import sys, json
for s in json.load(sys.stdin).get('result', []):
    print(f\"  {s.get('name','?')}: size={s.get('size',0)//1048576}MB created={s.get('creation_time','?')}\")
" 2>/dev/null || echo "No snapshots"
```

## Output Format

```
QDRANT ANALYSIS
===============
Version: [version] | Cluster: [status] | Peers: [count]
Collections: [count] | Total Points: [count]

ISSUES FOUND:
- [issue with affected collection/shard]

RECOMMENDATIONS:
- [actionable recommendation]
```

## Anti-Hallucination Rules

1. **NEVER assume resource names** — always discover via CLI/API in Phase 1 before referencing in Phase 2.
2. **NEVER fabricate metric names or dimensions** — verify against the service documentation or `--help` output.
3. **NEVER mix CLI commands between service versions** — confirm which version/API you are targeting.
4. **ALWAYS use the discovery → verify → analyze chain** — every resource referenced must have been discovered first.
5. **ALWAYS handle empty results gracefully** — an empty response is valid data, not an error to retry.

## Counter-Rationalizations

| Shortcut | Counter | Why |
|----------|---------|-----|
| "I'll skip discovery and check known resources" | Always run Phase 1 discovery first | Resource names change, new resources appear — assumed names cause errors |
| "The user only asked for a quick check" | Follow the full discovery → analysis flow | Quick checks miss critical issues; structured analysis catches silent failures |
| "Default configuration is probably fine" | Audit configuration explicitly | Defaults often leave logging, security, and optimization features disabled |
| "Metrics aren't needed for this" | Always check relevant metrics when available | API/CLI responses show current state; metrics reveal trends and intermittent issues |
| "I don't have access to that" | Try the command and report the actual error | Assumed permission failures prevent useful investigation; actual errors are informative |