cleveragents/cleveragents-core

Fork 3

UAT: Storage backends stay in-memory (no Tantivy/Qdrant/Neo4j/rdflib wiring) #7003

New issue

Open

opened 2026-04-10 06:23:19 +00:00 by HAL9000 · 1 comment

HAL9000 commented

2026-04-10 06:23:19 +00:00

Owner

What was tested

Reviewed the storage wiring in application/container.py and attempted to select the production backends declared in the Storage & Persistence spec (Tantivy text index, Qdrant vector DB, Neo4j graph, rdflib triple store).
Executed the vector backend factory with index.vector.backend=qdrant to confirm runtime behavior.

Expected behavior

Selecting tantivy, qdrant, or neo4j backends should load concrete implementations so the ACMS pipeline can persist text, vector, and graph data according to the spec.

Actual behavior

The DI container wires only InMemoryTextBackend, InMemoryVectorBackend, and InMemoryGraphBackend regardless of configuration. There are no Tantivy, Qdrant, Neo4j, or rdflib implementations anywhere in the codebase.
Running build_vector_backend with index.vector.backend=qdrant logs acms.vector_backend.fallback and returns InMemoryVectorBackend, confirming that the configurable “production” backend is missing.

Steps to reproduce

Inspect src/cleveragents/application/container.py lines 825-846: the providers always return the in-memory stub backends. There is no conditional wiring based on config.

Run:

import os, tempfile, sys
from pathlib import Path
sys.path.insert(0, "<repo>/src")
from cleveragents.config.settings import Settings
from cleveragents.infrastructure.database.unit_of_work import UnitOfWork
from cleveragents.application.services.vector_store_service import VectorStoreService
from cleveragents.application.services.faiss_vector_backend import build_vector_backend
work = Path(tempfile.mkdtemp())
os.environ["CLEVERAGENTS_DATA_DIR"] = str(work)
os.environ["CLEVERAGENTS_INDEX_VECTOR_BACKEND"] = "qdrant"
svc = VectorStoreService(Settings(), UnitOfWork(f"sqlite:///{work / 'cleveragents.db'}", require_confirmation=False))
backend = build_vector_backend(svc)
print(type(backend).__name__)

Output shows the warning acms.vector_backend.fallback configured_backend=qdrant fallback_backend=InMemoryVectorBackend and returns InMemoryVectorBackend.

Search the repository for Tantivy, Qdrant, Neo4j, or rdflib classes—only config strings exist; no usable adapters.

Code references

src/cleveragents/application/container.py (DI wiring)
src/cleveragents/application/services/faiss_vector_backend.py (only FAISS implementation with fallback)
src/cleveragents/domain/models/acms/index_stubs.py (in-memory stubs currently in use)

Impact

Storage & Persistence spec explicitly promises Tantivy (text), FAISS/Qdrant (vector), Neo4j/rdflib (graph). Without concrete adapters the ACMS indexing pipeline cannot satisfy those requirements, so server mode lacks the documented persistence guarantees.

## What was tested - Reviewed the storage wiring in `application/container.py` and attempted to select the production backends declared in the Storage & Persistence spec (Tantivy text index, Qdrant vector DB, Neo4j graph, rdflib triple store). - Executed the vector backend factory with `index.vector.backend=qdrant` to confirm runtime behavior. ## Expected behavior - Selecting `tantivy`, `qdrant`, or `neo4j` backends should load concrete implementations so the ACMS pipeline can persist text, vector, and graph data according to the spec. ## Actual behavior - The DI container wires only `InMemoryTextBackend`, `InMemoryVectorBackend`, and `InMemoryGraphBackend` regardless of configuration. There are no Tantivy, Qdrant, Neo4j, or rdflib implementations anywhere in the codebase. - Running `build_vector_backend` with `index.vector.backend=qdrant` logs `acms.vector_backend.fallback` and returns `InMemoryVectorBackend`, confirming that the configurable “production” backend is missing. ## Steps to reproduce 1. Inspect `src/cleveragents/application/container.py` lines 825-846: the providers always return the in-memory stub backends. There is no conditional wiring based on config. 2. Run: ```python import os, tempfile, sys from pathlib import Path sys.path.insert(0, "<repo>/src") from cleveragents.config.settings import Settings from cleveragents.infrastructure.database.unit_of_work import UnitOfWork from cleveragents.application.services.vector_store_service import VectorStoreService from cleveragents.application.services.faiss_vector_backend import build_vector_backend work = Path(tempfile.mkdtemp()) os.environ["CLEVERAGENTS_DATA_DIR"] = str(work) os.environ["CLEVERAGENTS_INDEX_VECTOR_BACKEND"] = "qdrant" svc = VectorStoreService(Settings(), UnitOfWork(f"sqlite:///{work / 'cleveragents.db'}", require_confirmation=False)) backend = build_vector_backend(svc) print(type(backend).__name__) ``` Output shows the warning `acms.vector_backend.fallback configured_backend=qdrant fallback_backend=InMemoryVectorBackend` and returns `InMemoryVectorBackend`. 3. Search the repository for `Tantivy`, `Qdrant`, `Neo4j`, or `rdflib` classes—only config strings exist; no usable adapters. ## Code references - `src/cleveragents/application/container.py` (DI wiring) - `src/cleveragents/application/services/faiss_vector_backend.py` (only FAISS implementation with fallback) - `src/cleveragents/domain/models/acms/index_stubs.py` (in-memory stubs currently in use) ## Impact - Storage & Persistence spec explicitly promises Tantivy (text), FAISS/Qdrant (vector), Neo4j/rdflib (graph). Without concrete adapters the ACMS indexing pipeline cannot satisfy those requirements, so server mode lacks the documented persistence guarantees.

HAL9000 added the

State

Unverified

label

2026-04-10 06:28:51 +00:00

HAL9000 referenced this issue

2026-04-10 06:33:48 +00:00

[AUTO-EPIC] Status: Epic Planner (Cycle 1) #6932

HAL9000 referenced this issue

2026-04-10 06:34:51 +00:00

[AUTO-ARCH] Architecture Review (Cycle 1) #6934

HAL9000 self-assigned this

2026-04-10 06:48:03 +00:00

HAL9000 added this to the v3.6.0 milestone

2026-04-10 06:48:03 +00:00

freemo added the

labels

2026-04-12 18:37:54 +00:00

HAL9000 added

and removed

labels

2026-04-14 07:19:40 +00:00

HAL9000 commented

2026-04-14 07:19:40 +00:00

Author

Owner

✅ Verified — UAT bug: storage backends stay in-memory — no real backend wiring. MoSCoW: Should-have. Priority: Medium.

Automated by CleverAgents Bot
Supervisor: Project Owner | Agent: project-owner-pool-supervisor

✅ **Verified** — UAT bug: storage backends stay in-memory — no real backend wiring. MoSCoW: Should-have. Priority: Medium. --- **Automated by CleverAgents Bot** Supervisor: Project Owner | Agent: project-owner-pool-supervisor