Skip to content

Navigation Menu

Sign in
Appearance settings
Sign in
Sign up
Appearance settings

Commit 6532806

Browse filesBrowse the repository at this point in the historyBrowse files
GoodbyePlanetclaude
GoodbyePlanet
and
claude
committed
feat: expose chunk_tier as a search filter
chunk_tier is computed for every indexed symbol and registered as a Qdrant keyword payload index, but no code path applied it as a filter, so clients couldn't scope a query to just classes or just methods. Adds chunk_tier to QdrantStore.search() / find_by_name() as a FieldCondition, and threads it through the search_code and find_symbol MCP tools. Closes #95 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
1 parent 65eea62 commit 6532806
Copy full SHA for 6532806

6 files changed

+122-14Lines changed: 122 additions & 14 deletions

File tree

Expand file treeCollapse file tree
Open diff view settings
Filter options
Expand file treeCollapse file tree
Open diff view settings
Collapse file

‎README.md‎

Copy file name to clipboardExpand all lines: README.md
+3-1Lines changed: 3 additions & 1 deletion
  • Display the source diff
  • Display the rich diff
Original file line numberDiff line numberDiff line change
@@ -360,7 +360,9 @@ in `.env`, then start without the `jina` profile (`docker-compose up` / `make do
360360

361361
`search_code` queries both via a Qdrant `query_points` call with `FusionQuery(fusion=RRF)`. Indexed
362362
payload fields (`language`, `service`, `symbol_type`, `chunk_tier`, `parent_name`, `file_path`) are
363-
usable as filters. The full payload also includes `signature`, `docstring`, `annotations`, `package`,
363+
usable as filters. `search_code` and `find_symbol` expose `chunk_tier` (`"method"` or `"class"`)
364+
directly, so a query can be scoped to just classes or just methods. The full payload also includes
365+
`signature`, `docstring`, `annotations`, `package`,
364366
`start_line`, `end_line`, `file_hash`, `indexed_at`, and language-specific extras (`http_method`,
365367
`http_route`, `spring_stereotype`, `lombok_annotations`, `is_async`, `uses_memo`, …).
366368

Collapse file

‎docs/retrieval-rrf.md‎

Copy file name to clipboardExpand all lines: docs/retrieval-rrf.md
+4-4Lines changed: 4 additions & 4 deletions
  • Display the source diff
  • Display the rich diff
Original file line numberDiff line numberDiff line change
@@ -100,24 +100,24 @@ semcode exposes four search tools to AI clients via the MCP protocol (`server/to
100100
### `search_code`
101101

102102
```
103-
search_code(query: str, service: str | None, limit: int = 10) -> str
103+
search_code(query: str, service: str | None, chunk_tier: str | None, limit: int = 10) -> str
104104
```
105105

106106
The primary semantic search tool. At query time:
107107

108108
1. Embeds the query string with both the dense provider (`embed_query`) and the sparse provider (`embed_query`)
109-
2. Calls `store.search()` with both vectors → RRF fusion
109+
2. Calls `store.search()` with both vectors → RRF fusion, optionally scoped to `chunk_tier` (`"method"` or `"class"`)
110110
3. Returns a formatted Markdown string with up to `limit` results
111111

112112
Each result includes: symbol name and type, RRF score, file location (path + line range), service, language, annotations, HTTP route (if present), and the symbol's signature or source (first 500 characters from the payload).
113113

114114
### `find_symbol`
115115

116116
```
117-
find_symbol(name: str, symbol_type: str | None, service: str | None, exact: bool = False) -> str
117+
find_symbol(name: str, symbol_type: str | None, service: str | None, chunk_tier: str | None, exact: bool = False) -> str
118118
```
119119

120-
Name-based lookup via `store.find_by_name()`. Does not use vectors or RRF. Returns up to 20 (exact) or 50 (substring) matches. Each result includes: name, type, location, package, parent class, and source (first 800 characters).
120+
Name-based lookup via `store.find_by_name()`. Does not use vectors or RRF. Supports filtering by `chunk_tier` (`"method"` or `"class"`) in addition to `symbol_type` and `service`. Returns up to 20 (exact) or 50 (substring) matches. Each result includes: name, type, location, package, parent class, and source (first 800 characters).
121121

122122
### `find_usages`
123123

Collapse file

‎server/store/qdrant.py‎

Copy file name to clipboardExpand all lines: server/store/qdrant.py
+13-6Lines changed: 13 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -216,14 +216,16 @@ async def search(
216216
sparse_vector: SparseVector,
217217
limit: int = 10,
218218
service: str | None = None,
219+
chunk_tier: str | None = None,
219220
) -> list[ScoredPoint]:
220-
query_filter = (
221-
Filter(
222-
must=[FieldCondition(key="service", match=MatchValue(value=service))]
221+
must = []
222+
if service:
223+
must.append(FieldCondition(key="service", match=MatchValue(value=service)))
224+
if chunk_tier:
225+
must.append(
226+
FieldCondition(key="chunk_tier", match=MatchValue(value=chunk_tier))
223227
)
224-
if service
225-
else None
226-
)
228+
query_filter = Filter(must=must) if must else None
227229

228230
result = await self._client.query_points(
229231
collection_name=self._collection,
@@ -252,6 +254,7 @@ async def find_by_name(
252254
name: str,
253255
symbol_type: str | None = None,
254256
service: str | None = None,
257+
chunk_tier: str | None = None,
255258
exact: bool = False,
256259
) -> list[ScoredPoint]:
257260
must = []
@@ -263,6 +266,10 @@ async def find_by_name(
263266
)
264267
if service:
265268
must.append(FieldCondition(key="service", match=MatchValue(value=service)))
269+
if chunk_tier:
270+
must.append(
271+
FieldCondition(key="chunk_tier", match=MatchValue(value=chunk_tier))
272+
)
266273

267274
base_filter = Filter(must=must) if must else None
268275

Collapse file

‎server/tools/search.py‎

Copy file name to clipboardExpand all lines: server/tools/search.py
+10-1Lines changed: 10 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -19,13 +19,15 @@ def register_search_tools(mcp: FastMCP) -> None:
1919
async def search_code(
2020
query: str,
2121
service: str | None = None,
22+
chunk_tier: str | None = None,
2223
limit: int = 10,
2324
) -> str:
2425
"""Semantically search code across indexed services using natural language.
2526
2627
Args:
2728
query: Natural language description of what you're looking for.
2829
service: Filter by service name
30+
chunk_tier: Filter by chunk tier: "method" or "class"
2931
limit: Maximum number of results (default 10)
3032
"""
3133
embedder = get_embedding_provider()
@@ -39,6 +41,7 @@ async def search_code(
3941
sparse_vector=sparse_vector,
4042
limit=limit,
4143
service=service,
44+
chunk_tier=chunk_tier,
4245
)
4346

4447
if not results:
@@ -74,6 +77,7 @@ async def find_symbol(
7477
name: str,
7578
symbol_type: str | None = None,
7679
service: str | None = None,
80+
chunk_tier: str | None = None,
7781
exact: bool = False,
7882
) -> str:
7983
"""Find a class, method, interface, or function by name.
@@ -82,11 +86,16 @@ async def find_symbol(
8286
name: Symbol name to search for
8387
symbol_type: Optional type filter: class, method, interface, enum, record, function, etc.
8488
service: Optional service filter
89+
chunk_tier: Optional chunk tier filter: "method" or "class"
8590
exact: If true, only exact name matches. If false (default), partial/fuzzy matching.
8691
"""
8792
store = get_store()
8893
results = await store.find_by_name(
89-
name=name, symbol_type=symbol_type, service=service, exact=exact
94+
name=name,
95+
symbol_type=symbol_type,
96+
service=service,
97+
chunk_tier=chunk_tier,
98+
exact=exact,
9099
)
91100

92101
if not results:
Collapse file

‎tests/test_store.py‎

Copy file name to clipboardExpand all lines: tests/test_store.py
+47-1Lines changed: 47 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,7 @@
44
from types import SimpleNamespace
55
from unittest.mock import AsyncMock, MagicMock
66

7-
from qdrant_client.models import Fusion, FusionQuery, SparseVector
7+
from qdrant_client.models import FieldCondition, Fusion, FusionQuery, SparseVector
88

99
from server.store.qdrant import QdrantStore
1010

@@ -99,3 +99,49 @@ async def test_search_uses_prefetch_and_rrf() -> None:
9999

100100
assert isinstance(kwargs["query"], FusionQuery)
101101
assert kwargs["query"].fusion == Fusion.RRF
102+
103+
104+
async def test_search_filters_by_chunk_tier() -> None:
105+
store = QdrantStore.__new__(QdrantStore)
106+
store._collection = "test"
107+
108+
fake_result = MagicMock()
109+
fake_result.points = []
110+
store._client = MagicMock()
111+
store._client.query_points = AsyncMock(return_value=fake_result)
112+
113+
dense = [0.1] * 768
114+
sparse = SparseVector(indices=[1, 2], values=[0.5, 0.3])
115+
116+
await store.search(
117+
dense_vector=dense, sparse_vector=sparse, limit=5, chunk_tier="method"
118+
)
119+
120+
kwargs = store._client.query_points.call_args.kwargs
121+
for prefetch in kwargs["prefetch"]:
122+
conditions = prefetch.filter.must
123+
assert any(
124+
isinstance(c, FieldCondition)
125+
and c.key == "chunk_tier"
126+
and c.match.value == "method"
127+
for c in conditions
128+
)
129+
130+
131+
async def test_find_by_name_filters_by_chunk_tier() -> None:
132+
store = QdrantStore.__new__(QdrantStore)
133+
store._collection = "test"
134+
135+
record = _make_record("MyService")
136+
store._client = MagicMock()
137+
store._client.scroll = AsyncMock(return_value=([record], None))
138+
139+
await store.find_by_name("MyService", exact=True, chunk_tier="class")
140+
141+
scroll_filter = store._client.scroll.call_args.kwargs["scroll_filter"]
142+
assert any(
143+
isinstance(c, FieldCondition)
144+
and c.key == "chunk_tier"
145+
and c.match.value == "class"
146+
for c in scroll_filter.must
147+
)
Collapse file

‎tests/tools/test_search.py‎

Copy file name to clipboardExpand all lines: tests/tools/test_search.py
+45-1Lines changed: 45 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -67,6 +67,46 @@ async def test_search_code_formats_hits() -> None:
6767
assert "0.870" in result
6868

6969

70+
async def test_search_code_passes_chunk_tier_to_store() -> None:
71+
search_code = _tool("search_code")
72+
store = AsyncMock()
73+
store.search.return_value = []
74+
75+
with (
76+
patch("server.tools.search.get_embedding_provider") as mock_embedder,
77+
patch("server.tools.search.get_sparse_provider") as mock_sparse,
78+
patch("server.tools.search.get_store", return_value=store),
79+
):
80+
mock_embedder.return_value.embed_query = AsyncMock(return_value=[0.1])
81+
mock_sparse.return_value.embed_query = AsyncMock(return_value={})
82+
await search_code("find the order service", chunk_tier="method")
83+
84+
store.search.assert_awaited_once_with(
85+
dense_vector=[0.1],
86+
sparse_vector={},
87+
limit=10,
88+
service=None,
89+
chunk_tier="method",
90+
)
91+
92+
93+
async def test_find_symbol_passes_chunk_tier_to_store() -> None:
94+
find_symbol = _tool("find_symbol")
95+
store = AsyncMock()
96+
store.find_by_name.return_value = []
97+
98+
with patch("server.tools.search.get_store", return_value=store):
99+
await find_symbol("OrderService", chunk_tier="class")
100+
101+
store.find_by_name.assert_awaited_once_with(
102+
name="OrderService",
103+
symbol_type=None,
104+
service=None,
105+
chunk_tier="class",
106+
exact=False,
107+
)
108+
109+
70110
async def test_find_symbol_reports_no_match() -> None:
71111
find_symbol = _tool("find_symbol")
72112
store = AsyncMock()
@@ -103,7 +143,11 @@ async def test_find_symbol_formats_match_with_parent() -> None:
103143
assert "`placeOrder`" in result
104144
assert "**Parent**: `OrderService`" in result
105145
store.find_by_name.assert_awaited_once_with(
106-
name="placeOrder", symbol_type=None, service=None, exact=True
146+
name="placeOrder",
147+
symbol_type=None,
148+
service=None,
149+
chunk_tier=None,
150+
exact=True,
107151
)
108152

109153

0 commit comments

Comments
0 (0)
Morty Proxy This is a proxified and sanitized view of the page, visit original site.