fix: enforce internal embeddings contract

This commit is contained in:
2026-08-08 17:33:19 +02:00
parent 2911e008d1
commit c875fa52ce
5 changed files with 131 additions and 3 deletions
@@ -90,3 +90,60 @@ git diff --check
- the focused harness verification still emits two pre-existing warnings: - the focused harness verification still emits two pre-existing warnings:
- `DeprecationWarning` from `testcontainers.postgres` - `DeprecationWarning` from `testcontainers.postgres`
- `FutureWarning` because `resources` currently flows through the legacy config translation path - `FutureWarning` because `resources` currently flows through the legacy config translation path
## Fix round 1 — 2026-08-08
### Findings addressed
- HIGH: external top-level `embeddings` remained an operational fallback and could still load
- MEDIUM: non-object embed JSON payloads escaped as raw `AttributeError`
### RED evidence
Command:
```bash
cd harness
./.venv/bin/pytest tests/test_internal_embeddings.py tests/test_config_resources.py tests/test_ollama_ensure.py -q
```
Observed before the fix:
- exit code `1`
- `2 failed, 36 passed, 2 warnings`
Representative failures:
- `AttributeError: 'list' object has no attribute 'get'` from `response.json()` returning a JSON array
- `Failed: DID NOT RAISE ConfigError` for top-level external `embeddings.provider=openai_compatible`
### GREEN evidence
Command:
```bash
cd harness
./.venv/bin/pytest tests/test_internal_embeddings.py tests/test_config_resources.py tests/test_ollama_ensure.py -q
```
Observed after the fix:
- exit code `0`
- `38 passed, 2 warnings`
Touched-file lint:
```bash
cd harness
./.venv/bin/ruff check tht/config.py tht/vectorstore/embeddings.py tests/test_internal_embeddings.py tests/test_config_resources.py
```
- exit code `0`
- `All checks passed!`
### Minimal fix
- validated the final active `cfg.embeddings` contract after config loading, so legacy top-level
embedding inputs now fail explicitly unless they exactly match the internal Ollama contract
- converted non-mapping embed JSON payloads into controlled `EmbeddingsError` failures with
sanitized diagnostics instead of raw attribute errors
+21 -2
View File
@@ -106,7 +106,7 @@ roots:
sessions: {runtime_root / 'sessions'} sessions: {runtime_root / 'sessions'}
artifacts: {runtime_root / 'artifacts'} artifacts: {runtime_root / 'artifacts'}
indexes: {runtime_root / 'indexes'} indexes: {runtime_root / 'indexes'}
embeddings: {{base_url: http://embedding.invalid, model: embed, dim: 768}} embeddings: {{provider: ollama_internal, base_url: http://embedding:11434, model: qwen3-embedding:0.6b, dim: 1024}}
""") """)
monkeypatch.setenv("THT_DATA_ROOT", str(data_root)) monkeypatch.setenv("THT_DATA_ROOT", str(data_root))
@@ -204,7 +204,7 @@ dwh:
vectors: vectors:
type: thoth_vector_http type: thoth_vector_http
writer: {base_url: https://vectors.test/, api_key: writer} writer: {base_url: https://vectors.test/, api_key: writer}
embeddings: {base_url: http://ollama:11434, dim: 768} embeddings: {provider: ollama_internal, base_url: http://embedding:11434, model: qwen3-embedding:0.6b, dim: 1024}
""" """
) )
@@ -302,6 +302,25 @@ dwh:
load_config(workspace) load_config(workspace)
def test_rejects_external_top_level_embedding_configuration(tmp_path):
workspace = tmp_path / "workspace.yaml"
workspace.write_text(
"""
dwh:
type: postgres_direct
connection: {database: analytics, schema: mart, user: reader, password: secret}
embeddings:
provider: openai_compatible
base_url: https://embedding.example.test
model: text-embedding-3-large
dim: 3072
"""
)
with pytest.raises(ConfigError, match="ollama_internal|provider|base_url|model|1024"):
load_config(workspace)
def test_builds_typed_evidence_sources_and_keeps_legacy_compatible(tmp_path): def test_builds_typed_evidence_sources_and_keeps_legacy_compatible(tmp_path):
common = """ common = """
dwh: dwh:
+17
View File
@@ -137,3 +137,20 @@ def test_internal_embeddings_reject_non_finite_values():
with pytest.raises(EmbeddingsError, match="finite|finit"): with pytest.raises(EmbeddingsError, match="finite|finit"):
embedder.embed(["alpha"]) embedder.embed(["alpha"])
def test_internal_embeddings_reject_non_object_json_payload():
from tht.vectorstore.embeddings import OllamaInternalEmbeddings
embedder = OllamaInternalEmbeddings(
EmbeddingsConfig(
provider="ollama_internal",
base_url="http://embedding:11434",
model="qwen3-embedding:0.6b",
dim=1024,
),
session=_Session([_Response([_vector(1.0)])]),
)
with pytest.raises(EmbeddingsError, match="response|payload|embeddings"):
embedder.embed(["alpha"])
+30
View File
@@ -454,6 +454,7 @@ def load_config(path: Path) -> Config:
if cfg.runtime_identity is not None if cfg.runtime_identity is not None
else path.resolve().as_posix() else path.resolve().as_posix()
) )
_validate_active_embeddings_config(cfg.embeddings, path)
return cfg return cfg
@@ -499,6 +500,35 @@ def _validate_internal_embedding_contract(raw: dict[str, Any], path: Path) -> No
) )
def _validate_active_embeddings_config(
embeddings: "EmbeddingsConfig | None",
path: Path,
) -> None:
if embeddings is None:
return
if embeddings.provider != "ollama_internal":
raise ConfigError(
f"Configurazione non valida in {path}:\n"
"embeddings.provider deve essere 'ollama_internal'"
)
if embeddings.model != "qwen3-embedding:0.6b":
raise ConfigError(
f"Configurazione non valida in {path}:\n"
"embeddings.model deve essere 'qwen3-embedding:0.6b'"
)
if embeddings.dim != 1024:
raise ConfigError(
f"Configurazione non valida in {path}:\n"
"embeddings.dim deve essere 1024"
)
if not _is_allowed_internal_embedding_url(embeddings.base_url):
raise ConfigError(
f"Configurazione non valida in {path}:\n"
"embeddings.base_url deve usare http://embedding:11434 "
"oppure un endpoint loopback di sviluppo su porta 11434"
)
def _is_allowed_internal_embedding_url(value: Any) -> bool: def _is_allowed_internal_embedding_url(value: Any) -> bool:
if not isinstance(value, str): if not isinstance(value, str):
return False return False
+6 -1
View File
@@ -33,7 +33,12 @@ class OllamaInternalEmbeddings:
raise EmbeddingsError( raise EmbeddingsError(
f"internal Ollama embeddings request failed for model {self.model}" f"internal Ollama embeddings request failed for model {self.model}"
) from exc ) from exc
payload = response.json() try:
payload = response.json()
except ValueError as exc:
raise EmbeddingsError("internal Ollama returned an invalid JSON response") from exc
if not isinstance(payload, dict):
raise EmbeddingsError("internal Ollama returned a non-object response payload")
embeddings = payload.get("embeddings") embeddings = payload.get("embeddings")
if not isinstance(embeddings, list): if not isinstance(embeddings, list):
raise EmbeddingsError("internal Ollama response is missing embeddings") raise EmbeddingsError("internal Ollama response is missing embeddings")