qdrant

Name: qdrant
Author: AlexAI-MCP

132stars24forks

How to install qdrant?

alexai-mcp/hermes-ccc/qdrant

$npx -y skills add AlexAI-MCP/hermes-CCC --skill qdrant

Installs into the current project.

Summary

High-performance vector search engine for production RAG — Rust-powered, horizontal scaling, hybrid dense+sparse search, metadata filtering.

Files · 1

View on GitHub

SKILL.md

1	# Qdrant — Production Vector Search Engine
2
3	High-performance, Rust-powered vector database for production RAG systems. Best for self-hosted deployments needing speed and horizontal scale.
4
5	## When to Use Qdrant vs Alternatives
6
7	- Qdrant: Production self-hosted, need speed + filtering + scale
8	- Chroma: Local dev, simple RAG prototypes
9	- Pinecone: Managed cloud, don't want to self-host
10	- FAISS: Pure in-memory, research, maximum speed
11
12	---
13
14	## Setup
15
16	```bash
17	pip install qdrant-client sentence-transformers
18
19	# Run Qdrant server
20	docker run -d -p 6333:6333 -p 6334:6334 \
21	-v $(pwd)/qdrant_storage:/qdrant/storage \
22	qdrant/qdrant
23	```
24
25	---
26
27	## Connect
28
29	```python
30	from qdrant_client import QdrantClient
31	from qdrant_client.models import Distance, VectorParams
32
33	# Local Docker
34	client = QdrantClient(host="localhost", port=6333)
35
36	# Cloud
37	client = QdrantClient(
38	url="https://your-cluster.aws.cloud.qdrant.io",
39	api_key="your-api-key"
40	)
41
42	# In-memory (testing)
43	client = QdrantClient(":memory:")
44	```
45
46	---
47
48	## Create Collection
49
50	```python
51	client.create_collection(
52	collection_name="my_docs",
53	vectors_config=VectorParams(
54	size=384, # match embedding model dimension
55	distance=Distance.COSINE, # COSINE \| EUCLID \| DOT
56	),
57	)
58	```
59
60	---
61
62	## Upsert Points
63
64	```python
65	from qdrant_client.models import PointStruct
66	from sentence_transformers import SentenceTransformer
67	import uuid
68
69	model = SentenceTransformer("all-MiniLM-L6-v2")
70
71	documents = [
72	{"text": "Python async programming", "source": "docs", "year": 2024},
73	{"text": "Machine learning with PyTorch", "source": "tutorial", "year": 2023},
74	]
75
76	embeddings = model.encode([d["text"] for d in documents])
77
78	points = [
79	PointStruct(
80	id=str(uuid.uuid4()),
81	vector=emb.tolist(),
82	payload=doc,
83	)
84	for doc, emb in zip(documents, embeddings)
85	]
86
87	client.upsert(collection_name="my_docs", points=points)
88	```
89
90	---
91
92	## Search
93
94	```python
95	from qdrant_client.models import Filter, FieldCondition, MatchValue
96
97	query = "how to write async code?"
98	q_vec = model.encode([query])[0].tolist()
99
100	# Basic search
101	results = client.search(
102	collection_name="my_docs",
103	query_vector=q_vec,
104	limit=5,
105	)
106
107	# With metadata filter
108	results = client.search(
109	collection_name="my_docs",
110	query_vector=q_vec,
111	query_filter=Filter(
112	must=[FieldCondition(key="source", match=MatchValue(value="docs"))]
113	),
114	limit=5,
115	with_payload=True,
116	)
117
118	for r in results:
119	print(f"[{r.score:.3f}] {r.payload['text']}")
120	```
121
122	---
123
124	## Hybrid Search (Dense + Sparse)
125
126	```python
127	from qdrant_client.models import SparseVector, NamedSparseVector, NamedVector
128
129	# Setup collection with both dense and sparse
130	client.create_collection(
131	collection_name="hybrid",
132	vectors_config={
133	"dense": VectorParams(size=384, distance=Distance.COSINE),
134	},
135	sparse_vectors_config={
136	"sparse": SparseVectorParams(),
137	},
138	)
139
140	# Search with RRF fusion
141	from qdrant_client.models import Prefetch, FusionQuery, Fusion
142
143	results = client.query_points(
144	collection_name="hybrid",
145	prefetch=[
146	Prefetch(query=dense_vec, using="dense", limit=20),
147	Prefetch(query=SparseVector(indices=[1,5,3], values=[0.1, 0.8, 0.5]),
148	using="sparse", limit=20),
149	],
150	query=FusionQuery(fusion=Fusion.RRF),
151	limit=5,
152	)
153	```
154
155	---
156
157	## Filtering Operations
158
159	```python
160	from qdrant_client.models import (
161	Filter, FieldCondition, MatchValue, MatchAny,
162	Range, HasIdCondition
163	)
164
165	# Match value
166	Filter(must=[FieldCondition(key="source", match=MatchValue(value="docs"))])
167
168	# Match any of
169	Filter(must=[FieldCondition(key="category", match=MatchAny(any=["tech", "science"]))])
170
171	# Range filter
172	Filter(must=[FieldCondition(key="year", range=Range(gte=2023, lte=2025))])
173
174	# Combine
175	Filter(
176	must=[FieldCondition(key="source", match=MatchValue(value="docs"))],
177	should=[FieldCondition(key="year", range=Range(gte=2024))],
178	must_not=[FieldCondition(key="archived", match=MatchValue(value=True))],
179	)
180	```
181
182	---
183
184	## Delete / Update
185
186	```python
187	# Delete by IDs
188	client.delete(collection_name="my_docs", points_selector=["id1", "id2"])
189
190	# Delete by filter
191	from qdrant_client.models import FilterSelector
192	client.delete(
193	collection_name="my_docs",
194	points_selector=FilterSelector(
195	filter=Filter(must=[FieldCondition(key="source", match=MatchValue(value="old"))])
196	)
197	)
198
199	# Collection info
200	info = client.get_collection("my_docs")
201	print(f"Vectors: {info.points_count}")
202	```
203
204	---
205
206	## Batch Upsert (Large Datasets)
207
208	```python
209	BATCH_SIZE = 100

People who liked this also viewed

6 picks

Type

find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 20262.6M27k
grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026633k182k
vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026573k29k
improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026522k182k
setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026454k182k
triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026424k182k

qdrant

132stars24forks

View on GitHub

How to install qdrant?

alexai-mcp/hermes-ccc/qdrant

$npx -y skills add AlexAI-MCP/hermes-CCC --skill qdrant

Installs into the current project.

Summary

High-performance vector search engine for production RAG — Rust-powered, horizontal scaling, hybrid dense+sparse search, metadata filtering.

Files · 1

View on GitHub

SKILL.md

1	# Qdrant — Production Vector Search Engine
2
3	High-performance, Rust-powered vector database for production RAG systems. Best for self-hosted deployments needing speed and horizontal scale.
4
5	## When to Use Qdrant vs Alternatives
6
7	- Qdrant: Production self-hosted, need speed + filtering + scale
8	- Chroma: Local dev, simple RAG prototypes
9	- Pinecone: Managed cloud, don't want to self-host
10	- FAISS: Pure in-memory, research, maximum speed
11
12	---
13
14	## Setup
15
16	```bash
17	pip install qdrant-client sentence-transformers
18
19	# Run Qdrant server
20	docker run -d -p 6333:6333 -p 6334:6334 \
21	-v $(pwd)/qdrant_storage:/qdrant/storage \
22	qdrant/qdrant
23	```
24
25	---
26
27	## Connect
28
29	```python
30	from qdrant_client import QdrantClient
31	from qdrant_client.models import Distance, VectorParams
32
33	# Local Docker
34	client = QdrantClient(host="localhost", port=6333)
35
36	# Cloud
37	client = QdrantClient(
38	url="https://your-cluster.aws.cloud.qdrant.io",
39	api_key="your-api-key"
40	)
41
42	# In-memory (testing)
43	client = QdrantClient(":memory:")
44	```
45
46	---
47
48	## Create Collection
49
50	```python
51	client.create_collection(
52	collection_name="my_docs",
53	vectors_config=VectorParams(
54	size=384, # match embedding model dimension
55	distance=Distance.COSINE, # COSINE \| EUCLID \| DOT
56	),
57	)
58	```
59
60	---
61
62	## Upsert Points
63
64	```python
65	from qdrant_client.models import PointStruct
66	from sentence_transformers import SentenceTransformer
67	import uuid
68
69	model = SentenceTransformer("all-MiniLM-L6-v2")
70
71	documents = [
72	{"text": "Python async programming", "source": "docs", "year": 2024},
73	{"text": "Machine learning with PyTorch", "source": "tutorial", "year": 2023},
74	]
75
76	embeddings = model.encode([d["text"] for d in documents])
77
78	points = [
79	PointStruct(
80	id=str(uuid.uuid4()),
81	vector=emb.tolist(),
82	payload=doc,
83	)
84	for doc, emb in zip(documents, embeddings)
85	]
86
87	client.upsert(collection_name="my_docs", points=points)
88	```
89
90	---
91
92	## Search
93
94	```python
95	from qdrant_client.models import Filter, FieldCondition, MatchValue
96
97	query = "how to write async code?"
98	q_vec = model.encode([query])[0].tolist()
99
100	# Basic search
101	results = client.search(
102	collection_name="my_docs",
103	query_vector=q_vec,
104	limit=5,
105	)
106
107	# With metadata filter
108	results = client.search(
109	collection_name="my_docs",
110	query_vector=q_vec,
111	query_filter=Filter(
112	must=[FieldCondition(key="source", match=MatchValue(value="docs"))]
113	),
114	limit=5,
115	with_payload=True,
116	)
117
118	for r in results:
119	print(f"[{r.score:.3f}] {r.payload['text']}")
120	```
121
122	---
123
124	## Hybrid Search (Dense + Sparse)
125
126	```python
127	from qdrant_client.models import SparseVector, NamedSparseVector, NamedVector
128
129	# Setup collection with both dense and sparse
130	client.create_collection(
131	collection_name="hybrid",
132	vectors_config={
133	"dense": VectorParams(size=384, distance=Distance.COSINE),
134	},
135	sparse_vectors_config={
136	"sparse": SparseVectorParams(),
137	},
138	)
139
140	# Search with RRF fusion
141	from qdrant_client.models import Prefetch, FusionQuery, Fusion
142
143	results = client.query_points(
144	collection_name="hybrid",
145	prefetch=[
146	Prefetch(query=dense_vec, using="dense", limit=20),
147	Prefetch(query=SparseVector(indices=[1,5,3], values=[0.1, 0.8, 0.5]),
148	using="sparse", limit=20),
149	],
150	query=FusionQuery(fusion=Fusion.RRF),
151	limit=5,
152	)
153	```
154
155	---
156
157	## Filtering Operations
158
159	```python
160	from qdrant_client.models import (
161	Filter, FieldCondition, MatchValue, MatchAny,
162	Range, HasIdCondition
163	)
164
165	# Match value
166	Filter(must=[FieldCondition(key="source", match=MatchValue(value="docs"))])
167
168	# Match any of
169	Filter(must=[FieldCondition(key="category", match=MatchAny(any=["tech", "science"]))])
170
171	# Range filter
172	Filter(must=[FieldCondition(key="year", range=Range(gte=2023, lte=2025))])
173
174	# Combine
175	Filter(
176	must=[FieldCondition(key="source", match=MatchValue(value="docs"))],
177	should=[FieldCondition(key="year", range=Range(gte=2024))],
178	must_not=[FieldCondition(key="archived", match=MatchValue(value=True))],
179	)
180	```
181
182	---
183
184	## Delete / Update
185
186	```python
187	# Delete by IDs
188	client.delete(collection_name="my_docs", points_selector=["id1", "id2"])
189
190	# Delete by filter
191	from qdrant_client.models import FilterSelector
192	client.delete(
193	collection_name="my_docs",
194	points_selector=FilterSelector(
195	filter=Filter(must=[FieldCondition(key="source", match=MatchValue(value="old"))])
196	)
197	)
198
199	# Collection info
200	info = client.get_collection("my_docs")
201	print(f"Vectors: {info.points_count}")
202	```
203
204	---
205
206	## Batch Upsert (Large Datasets)
207
208	```python
209	BATCH_SIZE = 100

People who liked this also viewed

6 picks

Type