Closes#14753
## What changed
| File | Change |
|---|---|
| `pyproject.toml` | `requires-python` → `>=3.13,<3.15`; remove
`strenum==0.4.15` |
| `Dockerfile` | `uv python install 3.13`, `uv sync --python 3.13` |
| `.github/workflows/tests.yml` | `uv sync --python 3.13` on both matrix
legs |
| `CLAUDE.md` | dev setup command + requirements note updated |
| `deepdoc/parser/mineru_parser.py` | `from strenum import StrEnum` →
`from enum import StrEnum` |
| `agent/tools/code_exec.py` | same |
`StrEnum` has been in the stdlib since Python 3.11 — the `strenum`
backport package is no longer needed once the floor is 3.13.
## Why uv.lock is not regenerated
`uv lock --python 3.13` fails because:
1. The infiniflow/graspologic fork pins `numpy>=1.26.4,<2.0.0`
2. `tensorflow-cpu>=2.20.0` (the first release with cp313 wheels)
depends on `ml-dtypes>=0.5.1`, which requires `numpy>=2.1.0`
3. These two constraints are irreconcilable on Python 3.13
The lockfile regeneration requires loosening the `numpy` upper bound in
the `infiniflow/graspologic` fork. Once that fork commit is updated and
the SHA in `pyproject.toml:49` is bumped, `uv lock --python 3.13` will
succeed.
## RFC corrections
Two claims in the original RFC (#14753) did not hold up under code
review:
- **"graspologic hard-blocks 3.13"** — the infiniflow fork at the pinned
commit has no `<3.13` Python constraint. The blocker is the transitive
`numpy<2.0.0` conflict with tensorflow-cpu's test dependency, not a
direct Python version cap.
- **"free-threading throughput gains for I/O-bound workload"** — Python
3.13 free-threading requires a special `--disable-gil` build and
provides no benefit for async I/O code (the GIL is already released
during I/O). The real motivation is forward compatibility and improved
error messages.
### What problem does this PR solve?
Visit
`http://127.0.0.1:9381/?__debugger__=yes&cmd=resource&f=debugger.js`
will expose the flask code:
```
docReady(() => {
if (!EVALEX_TRUSTED) {
initPinBox();
}
// if we are in console mode, show the console.
if (CONSOLE_MODE && EVALEX) {
createInteractiveConsole();
}
const frames = document.querySelectorAll("div.traceback div.frame");
if (EVALEX) {
addConsoleIconToFrames(frames);
}
addEventListenersToElements(document.querySelectorAll("div.detail"), "click", () =>
document.querySelector("div.traceback").scrollIntoView(false)
);
addToggleFrameTraceback(frames);
addToggleTraceTypesOnClick(document.querySelectorAll("h2.traceback"));
addInfoPrompt(document.querySelectorAll("span.nojavascript"));
wrapPlainTraceback();
});
function addToggleFrameTraceback(frames) {
frames.forEach((frame) => {
frame.addEventListener("click", () => {
frame.getElementsByTagName("pre")[0].parentElement.classList.toggle("expanded");
});
})
}
```
### Type of change
- [x] Other (please describe): Fix security risk
### What problem does this PR solve?
This PR fixes a mismatch between the MCP retrieval contract and the
backend retrieval API.
`ragflow_retrieval` already describes `dataset_ids` as optional, but
`/api/v1/retrieval` still rejected omitted or empty `dataset_ids` with
`` `dataset_ids` is required. ``. That made MCP retrieval fail even
though the tool schema promised that the request could search across all
available datasets.
This change updates `/api/v1/retrieval` to accept missing or empty
`dataset_ids`, resolve all accessible datasets for the authenticated
user, and keep the route schema aligned with the new runtime behavior.
It also adds focused unit coverage for the fallback resolution path and
the no-accessible-datasets case.
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
Fixes: #13981
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Improved dataset resolution to reliably discover all accessible
datasets through proper pagination, replacing the previous parsing
method.
* Enhanced error handling with clearer messaging when no datasets are
available for retrieval.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Fix the issue where the server-side parameter validation fails when the
id parameter is None in the asynchronous list_datasets method.
### What problem does this PR solve?
Fix the issue where the server-side parameter validation fails when the
id parameter is None in the asynchronous list_datasets method.
### Type of change
- [√ ] Bug Fix (non-breaking change which fixes an issue)
### What problem does this PR solve?
MCP host mode supports STREAMABLE-HTTP endpoint
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
The document metadata cache is built using the list documents endpoint
with default pagination parameters of page=1, page_size=3. This means
when using the MCP server to search a dataset, only chunks which come
from the first 30 documents in the dataset will have metadata returned.
Issue described in more detail here
https://github.com/infiniflow/ragflow/issues/11533
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
Co-authored-by: Giles Lloyd <giles.af.lloyd@gmail.com>
### What problem does this PR solve?
Expand the capabilities of the MCP Server. #8644.
Special thanks to @Drasek, this change is largely based on his original
implementation, it is super neat and well-structured to me. I basically
just integrated his code into the codebase with minimal modifications.
My main contribution is implementing a proper cache layer for dataset
and document metadata, using the LRU strategy with a 300s ± random 30s
TTL. The original code did not actually perform caching.
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
---------
Co-authored-by: Caspar Armster <caspar@armster.de>
### What problem does this PR solve?
Using the mcp server in n8n sometimes (with smaller models) results in
errors because the llm misses a char or adds one to the list of
dataset_ids provided. It first asks for the list of datasets and if you
got a larger list of them it makes a error recalling the list
completely. So adding the feature to just search through all available
datasets solves this and makes the retrieval of data more stable. The
functionality to just call special datasets by id is not changed, the
dataset_ids are now not required anymore (only the "question" is). You
can provide (like before) a list of datasets, a empty list or no list at
all.
### Type of change
- [X] New Feature (non-breaking change which adds functionality)
<img width="1897" height="880" alt="mcp error dataset id"
src="https://github.com/user-attachments/assets/71076d24-f875-4663-a69a-60839fc7a545"
/>
### What problem does this PR solve?
Enable MCP streamable-http model via docker compose
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Add MCP treamable-http transport.
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Add mcp self-host mode, a complement of #7084.
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
### What problem does this PR solve?
Add MCP support with a client example.
Issue link: #4344
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
