Merged
Conversation
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
5 issues found across 64 files
Prompt for AI agents (all issues)
Check if these issues are valid — if so, understand the root cause of each and fix them.
<file name="backend/onyx/tools/tool_implementations/file_reader/file_reader_tool.py">
<violation number="1" location="backend/onyx/tools/tool_implementations/file_reader/file_reader_tool.py:150">
P2: Validate start_char/num_chars and raise a ToolCallException on invalid numeric input instead of letting int() raise ValueError/TypeError.</violation>
</file>
<file name="backend/onyx/chat/llm_loop.py">
<violation number="1" location="backend/onyx/chat/llm_loop.py:221">
P1: The token budget calculation (lines 164-165 in original file) subtracts `project_files.total_token_count`, which presumably accounts for full file texts. When this fallback branch is taken to use `file_metadata_for_tool`, the actual message size may differ significantly from `total_token_count`.
1. If `total_token_count` is 0 (metadata-only mode), the budget is not reduced, risking context overflow when the metadata message is added.
2. If `total_token_count` includes full texts, the budget is over-reduced, causing unnecessary history truncation.
The budget calculation should be conditional on which project message type is actually used.</violation>
</file>
<file name="backend/tests/integration/tests/no_vectordb/conftest.py">
<violation number="1" location="backend/tests/integration/tests/no_vectordb/conftest.py:20">
P2: Add a timeout to the `requests.get` call so test collection doesn't hang indefinitely when the server is slow or unreachable.</violation>
</file>
<file name="backend/tests/integration/tests/no_vectordb/test_no_vectordb_chat.py">
<violation number="1" location="backend/tests/integration/tests/no_vectordb/test_no_vectordb_chat.py:125">
P2: The wait loop can exit without confirming the file was processed, so the test can proceed in a bad state and fail later. Add an explicit timeout assertion after the loop (or a flag) to fail fast when the file never becomes available.</violation>
</file>
<file name="backend/tests/unit/onyx/document_index/test_disabled_document_index.py">
<violation number="1" location="backend/tests/unit/onyx/document_index/test_disabled_document_index.py:68">
P1: Rule violated: **Use Pydantic over dataclasses**
Use Pydantic BaseModel for data modeling instead of Python dataclasses in new code (rule: Use Pydantic over dataclasses).</violation>
</file>
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
| from dataclasses import dataclass, field | ||
|
|
||
| # We only need a stub — the method raises before inspecting arguments. | ||
| @dataclass |
Contributor
There was a problem hiding this comment.
P1: Rule violated: Use Pydantic over dataclasses
Use Pydantic BaseModel for data modeling instead of Python dataclasses in new code (rule: Use Pydantic over dataclasses).
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At backend/tests/unit/onyx/document_index/test_disabled_document_index.py, line 68:
<comment>Use Pydantic BaseModel for data modeling instead of Python dataclasses in new code (rule: Use Pydantic over dataclasses).</comment>
<file context>
@@ -0,0 +1,191 @@
+ from dataclasses import dataclass, field
+
+ # We only need a stub — the method raises before inspecting arguments.
+ @dataclass
+ class _StubBatchParams:
+ doc_id_to_previous_chunk_cnt: dict[str, int] = field(default_factory=dict)
</file context>
|
|
||
| raw_file_id = cast(str, llm_kwargs[FILE_ID_FIELD]) | ||
| file_id = self._validate_file_id(raw_file_id) | ||
| start_char = max(0, int(llm_kwargs.get(START_CHAR_FIELD, 0))) |
Contributor
There was a problem hiding this comment.
P2: Validate start_char/num_chars and raise a ToolCallException on invalid numeric input instead of letting int() raise ValueError/TypeError.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At backend/onyx/tools/tool_implementations/file_reader/file_reader_tool.py, line 150:
<comment>Validate start_char/num_chars and raise a ToolCallException on invalid numeric input instead of letting int() raise ValueError/TypeError.</comment>
<file context>
@@ -0,0 +1,196 @@
+
+ raw_file_id = cast(str, llm_kwargs[FILE_ID_FIELD])
+ file_id = self._validate_file_id(raw_file_id)
+ start_char = max(0, int(llm_kwargs.get(START_CHAR_FIELD, 0)))
+ num_chars = min(
+ MAX_NUM_CHARS,
</file context>
| def _server_has_vector_db_disabled() -> bool: | ||
| """Query the running server to check whether DISABLE_VECTOR_DB is set.""" | ||
| try: | ||
| resp = requests.get( |
Contributor
There was a problem hiding this comment.
P2: Add a timeout to the requests.get call so test collection doesn't hang indefinitely when the server is slow or unreachable.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At backend/tests/integration/tests/no_vectordb/conftest.py, line 20:
<comment>Add a timeout to the `requests.get` call so test collection doesn't hang indefinitely when the server is slow or unreachable.</comment>
<file context>
@@ -0,0 +1,42 @@
+def _server_has_vector_db_disabled() -> bool:
+ """Query the running server to check whether DISABLE_VECTOR_DB is set."""
+ try:
+ resp = requests.get(
+ f"{API_SERVER_URL}/settings",
+ headers=GENERAL_HEADERS,
</file context>
Contributor
Greptile OverviewGreptile SummaryThis PR introduces a “no vector DB” deployment mode controlled by To preserve basic chat + project workflows without Vespa/OpenSearch, it adds a Confidence Score: 3/5
Important Files Changed
Sequence DiagramsequenceDiagram
participant User
participant API as API Server
participant Settings as /settings
participant Chat as Chat handler
participant Tools as construct_tools
participant FR as FileReaderTool
participant Store as FileStore
User->>API: GET /settings
API->>Settings: fetch_settings()
Settings-->>API: vector_db_enabled = !DISABLE_VECTOR_DB
API-->>User: settings JSON
User->>API: POST /chat/... message
API->>Chat: handle_stream_message_objects()
Chat->>Chat: collect available_file_ids (chat files + project files + persona user_files)
Chat->>Tools: construct_tools(file_reader_tool_config)
Tools-->>Chat: tools include FileReaderTool with allowed IDs
Chat->>API: run_llm_loop(...)
Note over User,FR: In no-vector-DB mode, LLM reads file sections on demand
User->>FR: tool call read_file(file_id, start_char, num_chars)
FR->>Store: load_user_file(file_id)
Store-->>FR: plaintext bytes (if available)
FR-->>User: section text response
|
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
Comment on lines
345
to
+369
|
|
||
| def _create_file_tool_metadata_message( | ||
| file_metadata: list[FileToolMetadata], | ||
| token_counter: Callable[[str], int], | ||
| ) -> ChatMessageSimple: | ||
| """Build a lightweight metadata-only message listing files available via FileReaderTool. | ||
|
|
||
| Used when files are too large to fit in context and the vector DB is | ||
| disabled, so the LLM must use ``read_file`` to inspect them. | ||
| """ | ||
| lines = [ | ||
| "You have access to the following files. Use the read_file tool to " | ||
| "read sections of any file:" | ||
| ] | ||
| for meta in file_metadata: | ||
| lines.append( | ||
| f'- {meta.file_id}: "{meta.filename}" (~{meta.approx_char_count:,} chars)' | ||
| ) | ||
|
|
||
| message_content = "\n".join(lines) | ||
| return ChatMessageSimple( | ||
| message=message_content, | ||
| token_count=token_counter(message_content), | ||
| message_type=MessageType.USER, | ||
| ) |
Contributor
There was a problem hiding this comment.
Wrong message role
_create_file_tool_metadata_message() returns a ChatMessageSimple with message_type=MessageType.USER. This injects synthetic metadata as a user message, which will be treated as user instruction/history rather than context (and can affect downstream logic that counts/filters user messages). This should be MessageType.SYSTEM (or whatever role you use for non-user context injections) to avoid contaminating conversation history.
Contributor
Additional Comments (1)
|
evan-onyx
force-pushed
the
feat/minimal-deployment-mode7
branch
from
ce9c51c to
edd4bc4
Compare
evan-onyx
force-pushed
the
feat/minimal-deployment-mode7
branch
from
edd4bc4 to
7fd8ded
Compare
evan-onyx
changed the title
Contributor
|
Preview Deployment
|
yuhongsun96
approved these changes
Feb 11, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Minimal version of Onyx. This deployment is primarily a chat frontend, with the ability to set up assistants that have access to manually uploaded files, but no connectors. This deployment method uses the following containers:
Notably, this does not include
launch a minimal onyx stack via docker-compose using
docker compose -f docker-compose.yml -f docker-compose.no-vectordb.yml -p $1 up -d --build
for helm you need to values a values override file with:
How Has This Been Tested?
tested in UI and added some unit tests for basic chat behavior in the presence of file reader tool
Additional Options
Summary by cubic
Adds a minimal deployment mode to run Onyx without a vector database. Core chat works and can read large files via a new File Reader tool; connectors and RAG search are disabled and related endpoints return 501.
New Features
Migration
Written for commit 527afa2. Summary will update on new commits.