refactor: rename the application package to docsgpt

The backend import package is now docsgpt, the name it will carry on PyPI; application was far too generic to install into anyone's site-packages. git mv plus a mechanical rewrite of every import, dotted string and path reference: 734 Python files, the compose files, Dockerfile, workflows, docs, setup scripts, devcontainer, k8s manifests, vscode config, pytest and coverage config, .gitignore. Behaviour is unchanged. Kept for one release: - A top-level application package whose meta-path finder resolves application.x.y to the already-imported docsgpt.x.y object, so old imports and entry points (celery -A application.app.celery, uvicorn application.asgi:asgi_app) keep working with a FutureWarning. - Celery registers every application.* task name as an alias of its docsgpt.* task on start-up, so messages queued by the previous release still run. The redbeat key prefix moves to redbeat:docsgpt:v2: so schedule entries the previous release wrote are left unread instead of firing twice. The backend image builds from the repository root (docker build -f docsgpt/Dockerfile .) so it can ship the alias package; a root .dockerignore allow-lists docsgpt/ and application/ and keeps caches, local data, .env files, the sample index files and the Dockerfile out. Compose and the image workflows point at the new context.
2026-10-13 10:15:21 +00:00 · 2026-09-07 10:20:43 +01:00
1 parent 3e7f1f91b4
commit 574f96341e
@@ -18,7 +18,7 @@ Then set in .env::
    DAYTONA_SNAPSHOT=docsgpt-artifacts-py312

 Keep the pins in sync with the backend venv (python-pptx / openpyxl / lxml /
-pillow are in application/requirements.txt; python-docx and reportlab arrive
+pillow are in docsgpt/requirements.txt; python-docx and reportlab arrive
 transitively) so the Daytona render output matches the Jupyter-backend output.
 """

@@ -57,7 +57,7 @@ def main(argv: list[str]) -> int:
    """Build (or skip) the snapshot and print the value to set as DAYTONA_SNAPSHOT."""
    args = _parse_args(argv)

-    from application.core.settings import settings
+    from docsgpt.core.settings import settings

    if not settings.DAYTONA_API_KEY:
        print("DAYTONA_API_KEY is not set (check .env).", file=sys.stderr)
@@ -14,7 +14,7 @@ only knobs operators need. This script discovers what's available from
 the :data:`BACKFILLERS` registry and runs whichever tables were asked for.

 This script imports ``pymongo`` directly. ``pymongo`` is not part of the
-base ``application/requirements.txt`` post-migration — install it
+base ``docsgpt/requirements.txt`` post-migration — install it
 directly before running::

    pip install 'pymongo>=4.6'
@@ -49,13 +49,13 @@ sys.path.insert(0, str(Path(__file__).resolve().parents[2]))

 from sqlalchemy import Connection, text  # noqa: E402

-from application.core.settings import settings  # noqa: E402
-from application.storage.db.engine import get_engine  # noqa: E402
+from docsgpt.core.settings import settings  # noqa: E402
+from docsgpt.storage.db.engine import get_engine  # noqa: E402


 # The backfill tool is the one remaining consumer of MongoDB in this repo.
 # It reads from Mongo and writes to Postgres, so it keeps its own client
-# rather than going through the (now-deleted) ``application.core.mongo_db``
+# rather than going through the (now-deleted) ``docsgpt.core.mongo_db``
 # wrapper. The DB name is hard-coded to ``docsgpt`` — historically surfaced
 # as ``settings.MONGO_DB_NAME`` but that setting has been removed post-cutover.
 _MONGO_DB_NAME = "docsgpt"
@@ -889,7 +889,7 @@ def _rename_faiss_indexes(
        )
        return stats

-    from application.storage.storage_creator import StorageCreator
+    from docsgpt.storage.storage_creator import StorageCreator

    storage = StorageCreator.get_storage()
    storage_type = getattr(storage, "__class__", type(storage)).__name__
@@ -1302,14 +1302,14 @@ def _backfill_memories(
    Mongo memory docs don't carry the body inline — ``content`` lives on
    disk at ``doc["storage_path"]`` (e.g.
    ``inputs/local/memories/<oid>/memory.txt``) and is accessed through
-    :class:`application.storage.storage_creator.StorageCreator`. We read
+    :class:`docsgpt.storage.storage_creator.StorageCreator`. We read
    the file lazily here so the PG ``content`` column gets the actual
    memory text rather than an empty string. Missing/unreadable files are
    logged and fall back to an empty ``content`` so one bad row doesn't
    abort the whole batch. Import is lazy (matches ``_rename_faiss_indexes``)
    so ``storage`` / backend creds aren't required to import this module.
    """
-    from application.storage.storage_creator import StorageCreator
+    from docsgpt.storage.storage_creator import StorageCreator

    storage = StorageCreator.get_storage()

@@ -1,7 +1,7 @@
 """Backfill ``token_usage.model_id`` for rows written before the column.

 New rows get ``model_id`` stamped at write time (see
-``application.llm.llm_creator`` / ``application.usage``). This script
+``docsgpt.llm.llm_creator`` / ``docsgpt.usage``). This script
 fills the historical NULLs by deriving the model from data we already
 trust, in priority order. A row is only ever filled by the
 highest-priority tier that matches it; tiers run in one transaction so
@@ -60,7 +60,7 @@ sys.path.insert(0, str(Path(__file__).resolve().parents[2]))

 from sqlalchemy import text  # noqa: E402

-from application.storage.db.engine import get_engine  # noqa: E402
+from docsgpt.storage.db.engine import get_engine  # noqa: E402


 # Tier 1: same request -> same model, primary (agent_stream) rows only.
@@ -40,7 +40,7 @@ sys.path.insert(0, str(Path(__file__).resolve().parents[2]))

 from sqlalchemy import text  # noqa: E402

-from application.storage.db.engine import get_engine  # noqa: E402
+from docsgpt.storage.db.engine import get_engine  # noqa: E402


 # Tier 1: parent message → user, conversation → agent.
@@ -13,7 +13,7 @@ Intended use:

 This script is a thin wrapper around ``alembic upgrade head``. It exists
 separately so the same command is discoverable from the repo root without
-remembering the ``-c application/alembic.ini`` invocation.
+remembering the ``-c docsgpt/alembic.ini`` invocation.
 """

 from __future__ import annotations
@@ -25,7 +25,7 @@ from alembic import command
 from alembic.config import Config

 REPO_ROOT = Path(__file__).resolve().parents[2]
-ALEMBIC_INI = REPO_ROOT / "application" / "alembic.ini"
+ALEMBIC_INI = REPO_ROOT / "docsgpt" / "alembic.ini"


 def main() -> int:
@@ -1,7 +1,7 @@
 """Rewrite stored model IDs across active config tables.

 Run this when a provider renames or deprecates a model ID. The catalog
-in ``application/core/models/<provider>.yaml`` is updated to the new ID,
+in ``docsgpt/core/models/<provider>.yaml`` is updated to the new ID,
 but existing agents and schedules still reference the old one and will
 fail on the next call. This script rewrites every active config row
 in-place inside a single transaction.
@@ -51,7 +51,7 @@ sys.path.insert(0, str(Path(__file__).resolve().parents[2]))

 from sqlalchemy import text  # noqa: E402

-from application.storage.db.session import db_session  # noqa: E402
+from docsgpt.storage.db.session import db_session  # noqa: E402


 # Built-in mapping reflects the 2026-05-25 Google preview -> GA swap.
@@ -28,7 +28,7 @@ verbatim as the conversation's prompt). See

 Run standalone (does NOT import anything from ``application/``). Python 3.11+.
 Flask is the only non-stdlib dependency and is already in
-``application/requirements.txt``.
+``docsgpt/requirements.txt``.

 Usage::

@@ -482,7 +482,7 @@ def _stream_chat_response(
    if tool_calls and toolcall_arg_mode:
        # Frame-split mode: the call's ``arguments`` are spread over several
        # deltas that all share one ``index``, which is what the client-side
-        # merge in application/llm/handlers/base.py reassembles.
+        # merge in docsgpt/llm/handlers/base.py reassembles.
        call = tool_calls[0]
        frames = _toolcall_arg_frames(call["function"]["arguments"], toolcall_arg_mode)
        for position, piece in enumerate(frames):
@@ -20,7 +20,7 @@ comes from ``MOCK_OIDC_GROUPS`` (comma-separated).

 Run standalone (does NOT import anything from ``application/``). Dependencies
 (flask, python-jose, cryptography, requests) are all in
-``application/requirements.txt``.
+``docsgpt/requirements.txt``.

 Usage::

@@ -255,7 +255,7 @@ FLASK_PID="$PIDDIR/flask.pid"
 log "starting Flask on 127.0.0.1:$FLASK_PORT"
 (
    cd "$E2E_TMP"
-    PYTHONUNBUFFERED=1 nohup "$FLASK_BIN" --app ../application/app.py run \
+    PYTHONUNBUFFERED=1 nohup "$FLASK_BIN" --app ../docsgpt/app.py run \
        --host 127.0.0.1 --port "$FLASK_PORT" \
        >"$FLASK_LOG" 2>&1 &
    echo $! > "$FLASK_PID"
@@ -275,7 +275,7 @@ log "starting Celery worker (solo pool)"
    cd "$E2E_TMP"
    PYTHONPATH="$REPO_ROOT${PYTHONPATH:+:$PYTHONPATH}" \
    PYTHONUNBUFFERED=1 \
-    nohup "$PY_BIN" -m celery -A application.app.celery worker \
+    nohup "$PY_BIN" -m celery -A docsgpt.app.celery worker \
        -l INFO --pool=solo -Q docsgpt,parsing \
        --without-gossip --without-mingle --without-heartbeat \
        >"$CELERY_LOG" 2>&1 &
@@ -290,7 +290,7 @@ if ! wait_for_log "celery 'celery@'" 30 "$CELERY_LOG" "celery@"; then
 fi

 # Ready check via `celery inspect ping`. We can't grep the log for 'ready'
-# because application/core/logging_config.py calls dictConfig with the default
+# because docsgpt/core/logging_config.py calls dictConfig with the default
 # disable_existing_loggers=True, which silences celery.worker's ready banner.
 # `inspect ping` queries the worker over the broker — it's the canonical
 # responsiveness check and doesn't depend on log output.
@@ -301,7 +301,7 @@ while (( elapsed < CELERY_INSPECT_TIMEOUT )); do
    if ( cd "$E2E_TMP" && \
         PYTHONPATH="$REPO_ROOT${PYTHONPATH:+:$PYTHONPATH}" \
         PYTHONUNBUFFERED=1 \
-         "$PY_BIN" -m celery -A application.app.celery inspect ping \
+         "$PY_BIN" -m celery -A docsgpt.app.celery inspect ping \
             --timeout 2 >/dev/null 2>&1 ); then
        ping_ok=1
        log "  -> celery inspect ping OK after ${elapsed}s"
@@ -9,7 +9,7 @@
 #   bash scripts/export_requirements.sh
 #
 # Each exported file is a complete environment (core plus the named extra),
-# so `pip install -r application/requirements-docling.txt` on its own works,
+# so `pip install -r docsgpt/requirements-docling.txt` on its own works,
 # and installing it on top of requirements.txt only adds the extra's packages.
 set -euo pipefail

@@ -40,18 +40,18 @@ export_file() {
  echo "wrote $out ($(grep -cE '^[A-Za-z0-9]' "$out") packages)"
 }

-export_file application/requirements.txt \
+export_file docsgpt/requirements.txt \
 "Core runtime. Optional extras live in requirements-<extra>.txt:
  docling  DOC_PARSER_ENGINE=docling, docling OCR backend, read_document structured output
  milvus   VECTOR_STORE=milvus"

-INDEX_URL=https://download.pytorch.org/whl/cpu export_file application/requirements-docling.txt \
+INDEX_URL=https://download.pytorch.org/whl/cpu export_file docsgpt/requirements-docling.txt \
 "Core runtime plus the docling extra: DOC_PARSER_ENGINE=docling, the docling
 OCR backend (layout-model hybrid OCR, ocrmac/rapidocr engines), .adoc/.vtt/.xml
 attachment parsing, and read_document's 'structured' output. The default
 anydoc engine needs none of this, and OCR itself does not either:
 OCR_ENABLED=true with the tesseract binary (or a DeepSeek-OCR endpoint) runs
-through application/parser/file/ocr_parser.py.
+through docsgpt/parser/file/ocr_parser.py.
 On Linux torch comes from the CPU-only PyTorch index (no CUDA stack); a GPU
 deployment can reinstall torch from PyPI on top.
 pip resolves the extra index as expected. uv only takes a package from the
@@ -61,7 +61,7 @@ pins the index per package) or set UV_INDEX_STRATEGY=unsafe-best-match.
 Docker: --build-arg EXTRAS=docling" \
  --extra docling

-export_file application/requirements-milvus.txt \
+export_file docsgpt/requirements-milvus.txt \
 "Core runtime plus the milvus extra (VECTOR_STORE=milvus): pymilvus and the
 embedded milvus-lite server, which pulls pyarrow.
 Docker: --build-arg EXTRAS=milvus" \
@@ -33,10 +33,10 @@ import dotenv  # noqa: E402

 dotenv.load_dotenv()

-from application.storage.db.repositories.auth_events import AuthEventsRepository  # noqa: E402
-from application.storage.db.repositories.user_roles import UserRolesRepository  # noqa: E402
-from application.storage.db.repositories.users import UsersRepository  # noqa: E402
-from application.storage.db.session import db_readonly, db_session  # noqa: E402
+from docsgpt.storage.db.repositories.auth_events import AuthEventsRepository  # noqa: E402
+from docsgpt.storage.db.repositories.user_roles import UserRolesRepository  # noqa: E402
+from docsgpt.storage.db.repositories.users import UsersRepository  # noqa: E402
+from docsgpt.storage.db.session import db_readonly, db_session  # noqa: E402

 logging.basicConfig(level=logging.WARNING, format="%(levelname)s %(message)s")
 logger = logging.getLogger("grant_admin")
@@ -52,8 +52,8 @@ def migrate_faiss_to_v1_vectorstore():

    vectors = list(vectors_collection.find())
    for vector in tqdm(vectors, desc="Migrating FAISS vectors"):
-        old_path = f"./application/indexes/{vector['user']}/{vector['name']}"
-        new_path = f"./application/indexes/{vector['_id']}"
+        old_path = f"./docsgpt/indexes/{vector['user']}/{vector['name']}"
+        new_path = f"./docsgpt/indexes/{vector['_id']}"
        try:
            os.makedirs(os.path.dirname(new_path), exist_ok=True)
            shutil.move(old_path, new_path)
@@ -54,7 +54,7 @@ from unittest.mock import MagicMock, patch


 # ---------------------------------------------------------------------------
-# Bootstrap: import application/* and connect to the configured Postgres.
+# Bootstrap: import docsgpt/* and connect to the configured Postgres.
 # ---------------------------------------------------------------------------

 ROOT = Path(__file__).resolve().parent.parent.parent
@@ -84,9 +84,9 @@ UNIQUE = f"qa_e2e_{int(time.time())}_{uuid.uuid4().hex[:6]}"

 # Quiet noisy library loggers so the script's own output stays readable.
 for name in (
-    "application", "application.api", "application.storage",
-    "application.usage", "application.parser",
-    "application.api.user.reconciliation",
+    "application", "docsgpt.api", "docsgpt.storage",
+    "docsgpt.usage", "docsgpt.parser",
+    "docsgpt.api.user.reconciliation",
 ):
    logging.getLogger(name).setLevel(logging.ERROR)

@@ -260,11 +260,11 @@ def s1_wal_and_reconciler() -> None:
    a stuck pending row to ``failed`` after 3 attempts and emits a
    structured alert."""

-    from application.api.answer.services.conversation_service import (
+    from docsgpt.api.answer.services.conversation_service import (
        ConversationService,
        TERMINATED_RESPONSE_PLACEHOLDER,
    )
-    from application.api.user.reconciliation import run_reconciliation
+    from docsgpt.api.user.reconciliation import run_reconciliation

    user = f"{UNIQUE}-s1"
    with ENGINE.begin() as c:
@@ -273,7 +273,7 @@ def s1_wal_and_reconciler() -> None:
            {"u": user},
        )

-    with patch_db_session("application.api.answer.services.conversation_service"):
+    with patch_db_session("docsgpt.api.answer.services.conversation_service"):
        res = ConversationService().save_user_question(
            conversation_id=None,
            question=f"{UNIQUE} s1 stuck WAL",
@@ -348,8 +348,8 @@ def s2_webhook_idempotency_and_scoping() -> None:

    from flask import Flask

-    from application.api.user.agents.webhooks import AgentWebhookListener
-    from application.storage.db.repositories.agents import AgentsRepository
+    from docsgpt.api.user.agents.webhooks import AgentWebhookListener
+    from docsgpt.storage.db.repositories.agents import AgentsRepository

    user_a = f"{UNIQUE}-s2a"
    user_b = f"{UNIQUE}-s2b"
@@ -382,8 +382,8 @@ def s2_webhook_idempotency_and_scoping() -> None:
    raw_key = f"{UNIQUE}-key-1"

    # Sequential dedup: same key twice → same task_id, single apply_async call.
-    with patch_db_session("application.api.user.agents.webhooks"), patch(
-        "application.api.user.agents.webhooks.process_agent_webhook.apply_async",
+    with patch_db_session("docsgpt.api.user.agents.webhooks"), patch(
+        "docsgpt.api.user.agents.webhooks.process_agent_webhook.apply_async",
        apply_mock,
    ):
        for _ in range(2):
@@ -410,8 +410,8 @@ def s2_webhook_idempotency_and_scoping() -> None:
    apply_mock.reset_mock()
    apply_calls.clear()
    cross_key = f"{UNIQUE}-cross"
-    with patch_db_session("application.api.user.agents.webhooks"), patch(
-        "application.api.user.agents.webhooks.process_agent_webhook.apply_async",
+    with patch_db_session("docsgpt.api.user.agents.webhooks"), patch(
+        "docsgpt.api.user.agents.webhooks.process_agent_webhook.apply_async",
        apply_mock,
    ):
        for ag in (agent_a, agent_b):
@@ -453,11 +453,11 @@ def s3_ingest_deterministic_and_resume() -> None:
    """Deterministic source_id is stable; chunk-progress checkpoint
    resumes from the next un-embedded chunk."""

-    from application.parser.embedding_pipeline import _read_resume_index
-    from application.storage.db.repositories.ingest_chunk_progress import (
+    from docsgpt.parser.embedding_pipeline import _read_resume_index
+    from docsgpt.storage.db.repositories.ingest_chunk_progress import (
        IngestChunkProgressRepository,
    )
-    from application.worker import _derive_source_id
+    from docsgpt.worker import _derive_source_id

    # Same scoped key → same uuid5; different scope → different uuid5.
    a = _derive_source_id("alice:UPLOAD-1")
@@ -475,7 +475,7 @@ def s3_ingest_deterministic_and_resume() -> None:

    # Checkpoint primitive: seed progress, observe resume index.
    src_id = uuid.uuid4()
-    with patch_db_session("application.parser.embedding_pipeline"):
+    with patch_db_session("docsgpt.parser.embedding_pipeline"):
        with ENGINE.begin() as conn:
            repo = IngestChunkProgressRepository(conn)
            repo.init_progress(str(src_id), total_chunks=10)
@@ -497,11 +497,11 @@ def s4_token_usage_atomic_rollback() -> None:
    insert; the surrounding transaction must roll back so no orphan
    token_usage row survives."""

-    from application.api.answer.services.conversation_service import (
+    from docsgpt.api.answer.services.conversation_service import (
        ConversationService,
    )
-    from application.storage.db.repositories import conversations as conv_mod
-    from application.storage.db.repositories.conversations import (
+    from docsgpt.storage.db.repositories import conversations as conv_mod
+    from docsgpt.storage.db.repositories.conversations import (
        ConversationsRepository,
    )

@@ -531,8 +531,8 @@ def s4_token_usage_atomic_rollback() -> None:
    try:
        conv_mod.ConversationsRepository.update_message_by_id = explode
        with patch_db_session(
-            "application.api.answer.services.conversation_service",
-        ), suppress_logging("application.api.answer.services.conversation_service"):
+            "docsgpt.api.answer.services.conversation_service",
+        ), suppress_logging("docsgpt.api.answer.services.conversation_service"):
            try:
                ConversationService().finalize_message(
                    mid,
@@ -577,7 +577,7 @@ def s5_reconciler_tool_call_sweeps() -> None:
    flips proposed→failed and executed→failed (or compensated when the
    tool exposes a working compensate)."""

-    from application.api.user.reconciliation import run_reconciliation
+    from docsgpt.api.user.reconciliation import run_reconciliation

    user = f"{UNIQUE}-s5"
    with ENGINE.begin() as c:
@@ -687,7 +687,7 @@ def s6_resume_janitor() -> None:
        )
    info(f"seeded pending_tool_state in 'resuming' (resumed_at=-11m) for conv={cid}")

-    from application.api.user.tasks import cleanup_pending_tool_state
+    from docsgpt.api.user.tasks import cleanup_pending_tool_state

    res = cleanup_pending_tool_state.run()
    info(f"janitor result: {res}")
@@ -712,13 +712,13 @@ def s7_token_usage_attribution() -> None:
    ``_persist_token_usage_inline`` self-write a row; the canary fires
    when ``status='complete'`` arrives with zero counts."""

-    from application.api.answer.services.conversation_service import (
+    from docsgpt.api.answer.services.conversation_service import (
        ConversationService,
    )
-    from application.storage.db.repositories.conversations import (
+    from docsgpt.storage.db.repositories.conversations import (
        ConversationsRepository,
    )
-    from application.usage import _maybe_persist_inline
+    from docsgpt.usage import _maybe_persist_inline

    user = f"{UNIQUE}-s7"
    with ENGINE.begin() as c:
@@ -747,7 +747,7 @@ def s7_token_usage_attribution() -> None:
        mid_zero = str(msg2["id"])

    # 1. Real counts → row written, source='agent_stream'.
-    with patch_db_session("application.api.answer.services.conversation_service"):
+    with patch_db_session("docsgpt.api.answer.services.conversation_service"):
        ok = ConversationService().finalize_message(
            mid_real,
            "answer",
@@ -771,7 +771,7 @@ def s7_token_usage_attribution() -> None:

    # 2. Zero-count complete fires the canary alert.
    canary_logger = logging.getLogger(
-        "application.api.answer.services.conversation_service"
+        "docsgpt.api.answer.services.conversation_service"
    )
    seen: list[logging.LogRecord] = []

@@ -787,7 +787,7 @@ def s7_token_usage_attribution() -> None:
    canary_logger.setLevel(logging.WARNING)
    try:
        with patch_db_session(
-            "application.api.answer.services.conversation_service",
+            "docsgpt.api.answer.services.conversation_service",
        ):
            ConversationService().finalize_message(
                mid_zero,
@@ -817,7 +817,7 @@ def s7_token_usage_attribution() -> None:
        _persist_token_usage_inline = True
        _token_usage_source = "compression"

-    with patch_db_session("application.usage"):
+    with patch_db_session("docsgpt.usage"):
        _maybe_persist_inline(
            _SideLLM(),
            {"prompt_tokens": 250, "generated_tokens": 80},
@@ -839,10 +839,10 @@ def s8_regenerate_replaces() -> None:
    position N before reserving the placeholder, so the new row lands at
    position=N (replacing the old) rather than appending at the end."""

-    from application.api.answer.services.conversation_service import (
+    from docsgpt.api.answer.services.conversation_service import (
        ConversationService,
    )
-    from application.storage.db.repositories.conversations import (
+    from docsgpt.storage.db.repositories.conversations import (
        ConversationsRepository,
    )

@@ -871,7 +871,7 @@ def s8_regenerate_replaces() -> None:
    assert before == [0, 1, 2, 3, 4], f"seed wrong: {before}"
    info(f"seeded conv={cid} positions={before}")

-    with patch_db_session("application.api.answer.services.conversation_service"):
+    with patch_db_session("docsgpt.api.answer.services.conversation_service"):
        result = ConversationService().save_user_question(
            conversation_id=cid,
            question=f"{UNIQUE}-regen",
@@ -899,7 +899,7 @@ def s9_idempotency_concurrent_claim() -> None:
    Confirms the ON CONFLICT DO NOTHING race semantics for the HTTP
    claim-first pattern."""

-    from application.storage.db.repositories.idempotency import (
+    from docsgpt.storage.db.repositories.idempotency import (
        IdempotencyRepository,
    )

@@ -1033,7 +1033,7 @@ def _build_openai_llm():
    """Construct an OpenAI-provider LLM pointing at the mock LLM. Skips
    the BYOM resolution path that LLMCreator usually goes through, since
    we just want a direct openai-client wrapper for the mock."""
-    from application.llm.openai import OpenAILLM
+    from docsgpt.llm.openai import OpenAILLM

    return OpenAILLM(
        api_key="mock-key",
@@ -1061,7 +1061,7 @@ def s11_live_llm_stream_through_wal() -> None:
      streamed response captured.
    """

-    from application.api.answer.services.conversation_service import (
+    from docsgpt.api.answer.services.conversation_service import (
        ConversationService,
    )

@@ -1072,7 +1072,7 @@ def s11_live_llm_stream_through_wal() -> None:
            {"u": user},
        )

-    with patch_db_session("application.api.answer.services.conversation_service"):
+    with patch_db_session("docsgpt.api.answer.services.conversation_service"):
        res = ConversationService().save_user_question(
            conversation_id=None,
            question=f"{UNIQUE} say hello",
@@ -1105,7 +1105,7 @@ def s11_live_llm_stream_through_wal() -> None:
        f"response[:40]={response_text[:40]!r}"
    )

-    with patch_db_session("application.api.answer.services.conversation_service"):
+    with patch_db_session("docsgpt.api.answer.services.conversation_service"):
        ok = ConversationService().finalize_message(
            mid,
            response_text,
@@ -1162,7 +1162,7 @@ def s12_full_embedding_loop() -> None:
    is just a regular Python function; ``self`` is a MagicMock with the
    minimum attrs the worker reads."""

-    from application.core.settings import settings
+    from docsgpt.core.settings import settings

    user = f"{UNIQUE}-s12"
    with ENGINE.begin() as c:
@@ -1227,7 +1227,7 @@ def s12_full_embedding_loop() -> None:
    settings.VECTOR_STORE = "faiss"

    idempotency_key = f"{user}:{UNIQUE}-s12-key"
-    from application.worker import _derive_source_id
+    from docsgpt.worker import _derive_source_id

    expected_source_id = str(_derive_source_id(idempotency_key))

@@ -1250,11 +1250,11 @@ def s12_full_embedding_loop() -> None:

    try:
        with patch_db_session(
-            "application.parser.embedding_pipeline",
-            "application.worker",
-            "application.api.answer.services.conversation_service",
-        ), patch("application.worker.upload_index", _fake_upload_index):
-            from application.worker import ingest_worker
+            "docsgpt.parser.embedding_pipeline",
+            "docsgpt.worker",
+            "docsgpt.api.answer.services.conversation_service",
+        ), patch("docsgpt.worker.upload_index", _fake_upload_index):
+            from docsgpt.worker import ingest_worker

            resp = ingest_worker(
                fake_self,
@@ -1347,7 +1347,7 @@ class CeleryWorkerService:
        self._proc = subprocess.Popen(  # noqa: S603
            [
                sys.executable, "-m", "celery",
-                "-A", "application.app.celery", "worker",
+                "-A", "docsgpt.app.celery", "worker",
                "-l", "INFO", "--pool=solo",
                "-Q", self._queue,
                "-n", f"qa-{UNIQUE}@%h",
@@ -1508,7 +1508,7 @@ def s13_sigkill_redelivers_via_acks_late() -> None:

    # Dispatch via celery's send_task using the same broker URL the worker
    # listens on. The DocsGPT celery instance has already loaded
-    # ``application.celeryconfig`` (broker = production .env DB 0); since
+    # ``docsgpt.celeryconfig`` (broker = production .env DB 0); since
    # Celery's conf re-reads ``CELERY_BROKER_URL`` from the env at publish
    # time, we override the env for the *whole* dispatch + wait block.
    from celery import Celery
@@ -1536,7 +1536,7 @@ def s13_sigkill_redelivers_via_acks_late() -> None:
        task_id = str(uuid.uuid4())
        info(f"dispatching task_id={task_id} key={idempotency_key}")
        chaos_app.send_task(
-            "application.api.user.tasks.process_agent_webhook",
+            "docsgpt.api.user.tasks.process_agent_webhook",
            kwargs={
                "agent_id": agent_id,
                "payload": {"event": f"{UNIQUE}-s13"},
@@ -1636,7 +1636,7 @@ def s10_celery_default_queue() -> None:
    """Sanity: the broker queue is project-scoped so a sibling worker on
    the same Redis can't grab DocsGPT tasks."""

-    from application import celeryconfig
+    from docsgpt import celeryconfig

    assert celeryconfig.task_default_queue == "docsgpt", (
        f"expected default queue 'docsgpt'; got {celeryconfig.task_default_queue!r}"