Skip to content

Commit 99008c8

Browse files
patelchaitanypatelchaitany
authored
feat: Added Agent skills for AI Agents (#6007)
Added Agent Skills for the feast user Signed-off-by: Chaitany patel <patelchaitany93@gmail.com>
1 parent d53a339 commit 99008c8

File tree

4 files changed

+1174
-0
lines changed

4 files changed

+1174
-0
lines changed

‎skills/SKILL.md‎

Lines changed: 253 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,253 @@
1+
---
2+
name: feast-user-guide
3+
description: Guide for working with Feast (Feature Store) — defining features, configuring feature_store.yaml, retrieving features online/offline, using the CLI, and building RAG retrieval pipelines. Use when the user asks about creating entities, feature views, on-demand feature views, stream feature views, feature services, data sources, feature_store.yaml configuration, feast apply/materialize commands, online or historical feature retrieval, or vector-based document retrieval with Feast.
4+
license: Apache-2.0
5+
compatibility: Works with Claude Code, OpenAI Codex, and any Agent Skills compatible tool.
6+
metadata:
7+
author: feast-dev
8+
version: "1.0"
9+
---
10+
11+
# Feast User Guide
12+
13+
## Quick Start
14+
15+
A Feast project requires:
16+
1. A `feature_store.yaml` config file
17+
2. Python files defining entities, data sources, feature views, and feature services
18+
3. Running `feast apply` to register definitions
19+
20+
```bash
21+
feast init my_project
22+
cd my_project
23+
feast apply
24+
```
25+
26+
## Core Concepts
27+
28+
### Entity
29+
An entity is a collection of semantically related features (e.g., a customer, a driver). Entities have join keys used to look up features.
30+
31+
```python
32+
from feast import Entity
33+
from feast.value_type import ValueType
34+
35+
driver = Entity(
36+
name="driver_id",
37+
description="Driver identifier",
38+
value_type=ValueType.INT64,
39+
)
40+
```
41+
42+
### Data Sources
43+
Data sources describe where raw feature data lives.
44+
45+
```python
46+
from feast import FileSource, BigQuerySource, KafkaSource, PushSource, RequestSource
47+
from feast.data_format import ParquetFormat
48+
49+
# Batch source (file)
50+
driver_stats_source = FileSource(
51+
name="driver_stats_source",
52+
path="data/driver_stats.parquet",
53+
timestamp_field="event_timestamp",
54+
created_timestamp_column="created",
55+
)
56+
57+
# Request source (for on-demand features)
58+
input_request = RequestSource(
59+
name="vals_to_add",
60+
schema=[Field(name="val_to_add", dtype=Float64)],
61+
)
62+
```
63+
64+
### FeatureView
65+
Maps features from a data source to entities with a schema, TTL, and online/offline settings.
66+
67+
```python
68+
from feast import FeatureView, Field
69+
from feast.types import Float32, Int64, String
70+
from datetime import timedelta
71+
72+
driver_hourly_stats = FeatureView(
73+
name="driver_hourly_stats",
74+
entities=[driver],
75+
ttl=timedelta(days=365),
76+
schema=[
77+
Field(name="conv_rate", dtype=Float32),
78+
Field(name="acc_rate", dtype=Float32),
79+
Field(name="avg_daily_trips", dtype=Int64),
80+
],
81+
online=True,
82+
source=driver_stats_source,
83+
)
84+
```
85+
86+
### OnDemandFeatureView
87+
Computes features at request time from other feature views and/or request data.
88+
89+
```python
90+
from feast import on_demand_feature_view
91+
import pandas as pd
92+
93+
@on_demand_feature_view(
94+
sources=[driver_hourly_stats, input_request],
95+
schema=[Field(name="conv_rate_plus_val", dtype=Float64)],
96+
mode="pandas",
97+
)
98+
def transformed_conv_rate(inputs: pd.DataFrame) -> pd.DataFrame:
99+
df = pd.DataFrame()
100+
df["conv_rate_plus_val"] = inputs["conv_rate"] + inputs["val_to_add"]
101+
return df
102+
```
103+
104+
### FeatureService
105+
Groups features from multiple views for retrieval.
106+
107+
```python
108+
from feast import FeatureService
109+
110+
driver_fs = FeatureService(
111+
name="driver_ranking",
112+
features=[driver_hourly_stats, transformed_conv_rate],
113+
)
114+
```
115+
116+
## Feature Retrieval
117+
118+
### Online (low-latency)
119+
```python
120+
from feast import FeatureStore
121+
122+
store = FeatureStore(repo_path=".")
123+
124+
features = store.get_online_features(
125+
features=[
126+
"driver_hourly_stats:conv_rate",
127+
"driver_hourly_stats:acc_rate",
128+
],
129+
entity_rows=[{"driver_id": 1001}, {"driver_id": 1002}],
130+
).to_dict()
131+
```
132+
133+
### Historical (training data with point-in-time joins)
134+
```python
135+
entity_df = pd.DataFrame({
136+
"driver_id": [1001, 1002],
137+
"event_timestamp": [datetime(2023, 1, 1), datetime(2023, 1, 2)],
138+
})
139+
140+
training_df = store.get_historical_features(
141+
entity_df=entity_df,
142+
features=["driver_hourly_stats:conv_rate", "driver_hourly_stats:acc_rate"],
143+
).to_df()
144+
```
145+
146+
Or use a FeatureService:
147+
```python
148+
training_df = store.get_historical_features(
149+
entity_df=entity_df,
150+
features=driver_fs,
151+
).to_df()
152+
```
153+
154+
## Materialization
155+
156+
Load features from offline store into online store:
157+
158+
```bash
159+
# Full materialization over a time range
160+
feast materialize 2023-01-01T00:00:00 2023-12-31T23:59:59
161+
162+
# Incremental (from last materialized timestamp)
163+
feast materialize-incremental $(date -u +"%Y-%m-%dT%H:%M:%S")
164+
```
165+
166+
Python API:
167+
```python
168+
from datetime import datetime
169+
store.materialize(start_date=datetime(2023, 1, 1), end_date=datetime(2023, 12, 31))
170+
store.materialize_incremental(end_date=datetime.utcnow())
171+
```
172+
173+
## CLI Commands
174+
175+
| Command | Purpose |
176+
|---------|---------|
177+
| `feast init [DIR]` | Create new feature repository |
178+
| `feast apply` | Register/update feature definitions |
179+
| `feast plan` | Preview changes without applying |
180+
| `feast materialize START END` | Materialize features to online store |
181+
| `feast materialize-incremental END` | Incremental materialization |
182+
| `feast entities list` | List registered entities |
183+
| `feast feature-views list` | List feature views |
184+
| `feast feature-services list` | List feature services |
185+
| `feast on-demand-feature-views list` | List on-demand feature views |
186+
| `feast teardown` | Remove infrastructure resources |
187+
| `feast version` | Show SDK version |
188+
189+
Options: `--chdir` / `-c` (run in different directory), `--feature-store-yaml` / `-f` (override config path).
190+
191+
## Vector Search / RAG
192+
193+
Define a feature view with vector fields for similarity search:
194+
195+
```python
196+
from feast.types import Array, Float32
197+
198+
wiki_passages = FeatureView(
199+
name="wiki_passages",
200+
entities=[passage_entity],
201+
schema=[
202+
Field(name="passage_text", dtype=String),
203+
Field(
204+
name="embedding",
205+
dtype=Array(Float32),
206+
vector_index=True,
207+
vector_length=384,
208+
vector_search_metric="COSINE",
209+
),
210+
],
211+
source=passages_source,
212+
online=True,
213+
)
214+
```
215+
216+
Retrieve similar documents:
217+
```python
218+
results = store.retrieve_online_documents(
219+
feature="wiki_passages:embedding",
220+
query=query_embedding,
221+
top_k=5,
222+
)
223+
```
224+
225+
## feature_store.yaml Minimal Config
226+
227+
```yaml
228+
project: my_project
229+
registry: data/registry.db
230+
provider: local
231+
online_store:
232+
type: sqlite
233+
path: data/online_store.db
234+
```
235+
236+
## Common Imports
237+
238+
```python
239+
from feast import (
240+
Entity, FeatureView, OnDemandFeatureView, FeatureService,
241+
Field, FileSource, RequestSource, FeatureStore,
242+
)
243+
from feast.on_demand_feature_view import on_demand_feature_view
244+
from feast.types import Float32, Float64, Int64, String, Bool, Array
245+
from feast.value_type import ValueType
246+
from datetime import timedelta
247+
```
248+
249+
## Detailed References
250+
251+
- **Feature definitions** (all types, parameters, patterns): See [references/feature-definitions.md](references/feature-definitions.md)
252+
- **Configuration** (feature_store.yaml, all store types, auth): See [references/configuration.md](references/configuration.md)
253+
- **Retrieval & RAG** (online/offline retrieval, vector search, RAG retriever): See [references/retrieval-and-rag.md](references/retrieval-and-rag.md)

0 commit comments

Comments
 (0)