[Bug]: 22:19:42 [ERROR][Exception]: Exceptions from Trio nursery (1 sub-exception) -- **ERROR**: POST predict: Post "http://127.0.0.1:33395/completion": EOF

[voycey]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (Language Policy).
• Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
• Please do not modify this template :) and fill in all the required fields.

RAGFlow workspace code commit ID

ge97fd2b5

RAGFlow image version

v0.19.x-40-ge97fd2b5 full

Other environment information

RTX A2000, Running on Docker

Actual behavior

Begin at:
Mon, 02 Jun 2025 18:14:20 GMT
Duration:
14728.60 s
Progress:
18:14:22 Task has been received.
18:15:27 Page(113): OCR started
18:15:45 Page(113): OCR finished (17.60s)
18:16:14 Page(113): Layout analysis (28.97s)
18:16:14 Page(113): Text extraction (29.04s)
18:16:28 Page(113): Start to generate keywords for every chunk ...
18:18:40 Page(113): Keywords generation 195 chunks completed in 132.48s
18:18:40 Page(113): Start to generate questions for every chunk ...
18:22:41 Page(113): Question generation 195 chunks completed in 240.64s
18:22:41 Page(113): Generate 195 chunks
18:22:44 Page(113): Embedding chunks (3.47s)
18:22:55 Page(113): Indexing done (10.29s). Task done (512.39s)
18:14:41 Task has been received.
18:14:44 Page(1325): OCR started
18:15:00 Page(1325): OCR finished (16.73s)
18:15:23 Page(1325): Layout analysis (22.43s)
18:15:23 Page(1325): Text extraction (22.79s)
18:15:35 Page(1325): Start to generate keywords for every chunk ...
18:16:17 Page(1325): Keywords generation 230 chunks completed in 41.94s
18:16:17 Page(1325): Start to generate questions for every chunk ...
18:18:09 Page(1325): Question generation 230 chunks completed in 112.58s
18:18:10 Page(1325): Generate 230 chunks
18:18:14 Page(1325): Embedding chunks (4.85s)
18:18:37 Page(1325): Indexing done (22.64s). Task done (236.46s)
18:14:41 Task has been received.
18:16:17 Page(6173): OCR started
18:16:27 Page(6173): OCR finished (10.33s)
18:16:54 Page(6173): Layout analysis (26.22s)
18:16:54 Page(6173): Text extraction (26.31s)
18:17:10 Page(6173): Start to generate keywords for every chunk ...
18:19:31 Page(6173): Keywords generation 331 chunks completed in 140.61s
18:19:31 Page(6173): Start to generate questions for every chunk ...
18:26:06 Page(6173): Question generation 331 chunks completed in 395.17s
18:26:06 Page(6173): Generate 331 chunks
18:26:12 Page(6173): Embedding chunks (5.52s)
18:26:28 Page(6173): Indexing done (16.69s). Task done (707.03s)
18:14:41 Task has been received.
18:16:59 Page(3749): OCR started
18:17:09 Page(3749): OCR finished (10.06s)
18:17:35 Page(3749): Layout analysis (25.91s)
18:17:35 Page(3749): Text extraction (25.96s)
18:17:51 Page(3749): Start to generate keywords for every chunk ...
18:20:24 Page(3749): Keywords generation 375 chunks completed in 152.80s
18:20:24 Page(3749): Start to generate questions for every chunk ...
18:29:48 Page(3749): Question generation 375 chunks completed in 563.96s
18:29:49 Page(3749): Generate 375 chunks
18:29:55 Page(3749): Embedding chunks (6.57s)
18:30:51 Page(3749): Indexing done (55.81s). Task done (970.12s)
18:14:41 Task has been received.
18:17:40 Page(2537): OCR started
18:17:49 Page(2537): OCR finished (8.79s)
18:18:13 Page(2537): Layout analysis (24.25s)
18:18:13 Page(2537): Text extraction (24.38s)
18:18:31 Page(2537): Start to generate keywords for every chunk ...
18:21:09 Page(2537): Keywords generation 261 chunks completed in 158.01s
18:21:09 Page(2537): Start to generate questions for every chunk ...
18:33:44 Page(2537): Question generation 261 chunks completed in 754.62s
18:33:44 Page(2537): Generate 261 chunks
18:33:49 Page(2537): Embedding chunks (4.59s)
18:34:02 Page(2537): Indexing done (13.68s). Task done (1161.46s)
18:14:41 Task has been received.
18:18:18 Page(4961): OCR started
18:18:29 Page(4961): OCR finished (10.13s)
18:18:55 Page(4961): Layout analysis (26.05s)
18:18:55 Page(4961): Text extraction (26.34s)
18:19:10 Page(4961): Start to generate keywords for every chunk ...
18:23:30 Page(4961): Keywords generation 323 chunks completed in 260.04s
18:23:30 Page(4961): Start to generate questions for every chunk ...
18:36:55 Page(4961): Question generation 323 chunks completed in 805.44s
18:36:55 Page(4961): Generate 323 chunks
18:37:01 Page(4961): Embedding chunks (5.55s)
18:37:16 Page(4961): Indexing done (15.19s). Task done (1354.89s)
18:14:41 Task has been received.
18:18:59 Page(7385): OCR started
18:19:10 Page(7385): OCR finished (10.63s)
18:19:34 Page(7385): Layout analysis (23.97s)
18:19:34 Page(7385): Text extraction (24.24s)
18:19:50 Page(7385): Start to generate keywords for every chunk ...
18:27:07 Page(7385): Keywords generation 330 chunks completed in 437.78s
18:27:08 Page(7385): Start to generate questions for every chunk ...
18:40:08 Page(7385): Question generation 330 chunks completed in 780.08s
18:40:08 Page(7385): Generate 330 chunks
18:40:13 Page(7385): Embedding chunks (5.45s)
18:40:29 Page(7385): Indexing done (15.30s). Task done (1547.09s)
18:14:42 Task has been received.
18:19:39 Page(97109): OCR started
18:19:48 Page(97109): OCR finished (9.00s)
18:20:13 Page(97109): Layout analysis (24.64s)
18:20:13 Page(97109): Text extraction (24.89s)
18:20:27 Page(97109): Start to generate keywords for every chunk ...
18:30:48 Page(97109): Keywords generation 295 chunks completed in 620.71s
18:30:48 Page(97109): Start to generate questions for every chunk ...
18:44:47 Page(97109): Question generation 295 chunks completed in 839.32s
18:44:47 Page(97109): Generate 295 chunks
18:44:53 Page(97109): Embedding chunks (5.15s)
18:45:09 Page(97109): Indexing done (16.60s). Task done (1827.16s)
18:14:42 Task has been received.
18:20:17 Page(8597): OCR started
18:20:27 Page(8597): OCR finished (9.62s)
18:20:53 Page(8597): Layout analysis (25.65s)
18:20:53 Page(8597): Text extraction (25.72s)
18:21:04 Page(8597): Start to generate keywords for every chunk ...
18:31:33 Page(8597): Keywords generation 264 chunks completed in 628.17s
18:31:33 Page(8597): Start to generate questions for every chunk ...
18:46:55 Page(8597): Question generation 264 chunks completed in 922.41s
18:46:55 Page(8597): Generate 264 chunks
18:47:00 Page(8597): Embedding chunks (4.25s)
18:47:14 Page(8597): Indexing done (14.50s). Task done (1952.58s)
18:14:42 Task has been received.
18:20:57 Page(109121): OCR started
18:21:14 Page(109121): OCR finished (17.02s)
18:21:37 Page(109121): Layout analysis (22.94s)
18:21:37 Page(109121): Text extraction (23.25s)
18:21:49 Page(109121): Start to generate keywords for every chunk ...
18:34:24 Page(109121): Keywords generation 321 chunks completed in 755.37s
18:34:24 Page(109121): Start to generate questions for every chunk ...
18:51:43 Page(109121): Question generation 321 chunks completed in 1038.71s
18:51:43 Page(109121): Generate 321 chunks
18:51:48 Page(109121): Embedding chunks (4.87s)
18:52:03 Page(109121): Indexing done (14.66s). Task done (2240.08s)
18:18:37 Task has been received.
18:21:40 Page(121133): OCR started
18:21:54 Page(121133): OCR finished (14.25s)
18:22:18 Page(121133): Layout analysis (23.78s)
18:22:18 Page(121133): Text extraction (23.86s)
18:22:21 Page(121133): Start to generate keywords for every chunk ...
18:34:33 Page(121133): Keywords generation 37 chunks completed in 731.82s
18:34:33 Page(121133): Start to generate questions for every chunk ...
18:52:06 Page(121133): Question generation 37 chunks completed in 1053.28s
18:52:07 Page(121133): Generate 37 chunks
18:52:09 Page(121133): Embedding chunks (1.13s)
18:52:11 Page(121133): Indexing done (1.96s). Task done (2013.42s)
18:22:55 Task has been received.
18:22:55 Page(133145): OCR started
18:23:03 Page(133145): OCR finished (7.54s)
18:23:28 Page(133145): Layout analysis (24.81s)
18:23:28 Page(133145): Text extraction (25.19s)
18:23:38 Page(133145): Start to generate keywords for every chunk ...
18:37:28 Page(133145): Keywords generation 259 chunks completed in 829.47s
18:37:28 Page(133145): Start to generate questions for every chunk ...
18:54:17 Page(133145): Question generation 259 chunks completed in 1008.87s
18:54:17 Page(133145): Generate 259 chunks
18:54:23 Page(133145): Embedding chunks (5.61s)
18:54:35 Page(133145): Indexing done (12.14s). Task done (1900.16s)
18:26:29 Task has been received.
18:26:29 Page(145157): OCR started
18:27:00 Page(145157): OCR finished (31.00s)
18:27:24 Page(145157): Layout analysis (24.42s)
18:27:24 Page(145157): Text extraction (24.55s)
18:27:45 Page(145157): Start to generate keywords for every chunk ...
18:42:28 Page(145157): Keywords generation 874 chunks completed in 882.74s
18:42:28 Page(145157): Start to generate questions for every chunk ...
19:02:22 Page(145157): Question generation 874 chunks completed in 1193.92s
19:02:22 Page(145157): Generate 874 chunks
19:02:36 Page(145157): Embedding chunks (13.75s)
19:03:18 Page(145157): Indexing done (41.56s). Task done (2209.15s)
18:30:51 Task has been received.
18:30:52 Page(157167): OCR started
18:31:07 Page(157167): OCR finished (15.19s)
18:31:26 Page(157167): Layout analysis (18.98s)
18:31:26 Page(157167): Text extraction (19.08s)
18:31:54 Page(157167): Start to generate keywords for every chunk ...
18:49:47 Page(157167): Keywords generation 1376 chunks completed in 1072.87s
18:49:48 Page(157167): Start to generate questions for every chunk ...
19:12:30 Page(157167): Question generation 1376 chunks completed in 1362.05s
19:12:30 Page(157167): Generate 1376 chunks
19:12:50 Page(157167): Embedding chunks (19.77s)
19:13:46 Page(157167): Indexing done (56.12s). Task done (2574.18s)
19:13:48 created task raptor
21:52:11 Task has been received.
22:19:42 [ERROR][Exception]: Exceptions from Trio nursery (1 sub-exception) -- ERROR: POST predict: Post "http://127.0.0.1:33395/completion": EOF

Seemed to work ok but couldnt send the notification to the completion endpoint? Some Docker servers dont accept 127.0.0.1 due to networking - this should be renamed to be the ragflow-server container I guess?

What is the payload for Parse completion so I can mark these as complete manually via curl?

Expected behavior

Parsing should mark the document as complete

Steps to reproduce

Simple PDF Parse with RAPTOR only

Additional information

No response

[dosubot]

The EOF error when posting to http://127.0.0.1:33395/completion is likely due to Docker networking: 127.0.0.1 inside a container refers to the container itself, not the host or other containers. For inter-container communication, use the Docker Compose service name (e.g., ragflow-server) or the container's network alias instead of 127.0.0.1. The docker-compose.yml sets up a ragflow-server container and a custom bridge network, so you should reference endpoints as http://ragflow-server:PORT/… rather than localhost or 127.0.0.1. You can also use extra_hosts to map host.docker.internal if you need to reach the host from a container, but for container-to-container calls, always use service names. See the relevant configuration and networking details here.

For marking a document or chunk as complete via curl, the API typically expects a JSON payload with a status or availability field. For example, to update a chunk's availability, use:

curl -X PUT \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <YOUR_TOKEN>" \
  -d '{"available": true}' \
  http://ragflow-server:9380/api/v1/datasets/<DATASET_ID>/chunks/<CHUNK_ID>

The parameter name is available and it accepts boolean (true/false) or numeric (1/0) values. For document status, the payload might be {"enable": false} or similar, depending on the endpoint. See examples here and here.

If you need to mark a parse as complete, you will need the correct endpoint and IDs for your document or chunk. If you are unsure of the exact endpoint, check your API documentation or inspect the logs for the endpoint being called.

If you continue to see EOF errors, double-check that all services are on the same Docker network and that you are not using 127.0.0.1 for cross-container calls. Update your configuration or scripts to use the ragflow-server service name as the host.

To reply, just mention my name @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}   

[voycey]

@dosu how do I mark an etnire document as complete? It seems that is where it is failing

[dosubot]

To mark an entire document as complete in RAGFlow, you need to update its status using the API. The relevant endpoint is /change_status, which requires a POST request with the following JSON payload:

{
  "doc_id": "<YOUR_DOCUMENT_ID>",
  "status": "1"
}

Here, "status": "1" marks the document as enabled/active (and "0" disables it). This endpoint checks your authorization and updates the document's status in the database and document store. Make sure to use the correct document ID.

Example curl command (replace placeholders as needed):

curl -X POST \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <YOUR_TOKEN>" \
  -d '{"doc_id": "<YOUR_DOCUMENT_ID>", "status": "1"}' \
  http://ragflow-server:9380/api/document/change_status

If you are running inside Docker, ensure you use the service name (ragflow-server) instead of 127.0.0.1 for the host in your URL, as Docker networking does not route 127.0.0.1 between containers reference.

If you want to disable a document (make it unavailable), use "status": "0" in the payload.

If you need to mark the document as logically "complete" (i.e., all tasks finished), this is managed internally by the backend when all associated tasks are done. If your document is stuck, it may be due to a task or progress tracking issue. In that case, check the document's progress and task status in the database or via the API reference.

Let me know if you need help finding your document ID or further troubleshooting the completion logic.

To reply, just mention my name @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}   

[voycey]

@dosubot

ERROR    22 404 Not Found: The requested URL was not found on the server. If you entered the URL manually please check your spelling and try again.
Traceback (most recent call last):
  File "/ragflow/.venv/lib/python3.10/site-packages/flask/app.py", line 880, in full_dispatch_request
    rv = self.dispatch_request()
  File "/ragflow/.venv/lib/python3.10/site-packages/flask/app.py", line 854, in dispatch_request
    self.raise_routing_exception(req)
  File "/ragflow/.venv/lib/python3.10/site-packages/flask/app.py", line 463, in raise_routing_exception
    raise request.routing_exception  # type: ignore[misc]
  File "/ragflow/.venv/lib/python3.10/site-packages/flask/ctx.py", line 362, in match_request
    result = self.url_adapter.match(return_rule=True)  # type: ignore
  File "/ragflow/.venv/lib/python3.10/site-packages/werkzeug/routing/map.py", line 629, in match
    raise NotFound() from None
werkzeug.exceptions.NotFound: 404 Not Found: The requested URL was not found on the server. If you entered the URL manually please check your spelling and try again.

How can I update this in the databas or elasticsearch instead?

[dosubot]

You can update the document status directly in the database or Elasticsearch, bypassing the API.

For the database:
The document status is stored in the document table as the status field (CharField, "1" = valid/active, "0" = wasted). You can run a SQL query like:

UPDATE document SET status = '1' WHERE id = '<YOUR_DOCUMENT_ID>';

If you also want to mark parsing as complete, you may need to update the run field to the value representing "done" (typically "3" for DONE), and set progress to 1.0:

UPDATE document SET status = '1', run = '3', progress = 1.0 WHERE id = '<YOUR_DOCUMENT_ID>';

Reference: Document model schema

For Elasticsearch:
If you use ES as your document store, you can update the status using an UpdateByQuery or direct update. The ESConnection class in RAGFlow does this by matching on id and kb_id, then setting the status field:

POST /<your_index>/_update_by_query
{
  "script": {
    "source": "ctx._source.status = params.status;",
    "params": {
      "status": "1"
    }
  },
  "query": {
    "term": {
      "id": "<YOUR_DOCUMENT_ID>"
    }
  }
}

Reference: ESConnection update logic

Make sure to replace <YOUR_DOCUMENT_ID> and <your_index> with your actual values. If you need to update both the database and Elasticsearch, do both steps to keep them in sync. If you need the exact value for "done" in the run field, check your TaskStatus enum or the document_service.py logic.

Let me know if you need help with the exact SQL for your DB engine or the ES curl command.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}   

[voycey]

@dosu - are there any other tasks that this might miss just marking it as complete?
The failure message was that the POST to completion failed:

ERROR: POST predict: Post "http://127.0.0.1:36617/completion"

[dosubot]

If you manually mark a document as complete in the database and Elasticsearch, you generally cover all the core state and index updates—RAGFlow’s backend is designed so that document completion is handled synchronously and self-contained within the workflow (updating DB, ES, cleaning up storage, updating stats). There aren’t any automatic background jobs, notifications, or hooks triggered just by marking a document as complete—so you won’t miss any internal RAGFlow processes by doing it this way.

However, if you have any external orchestration, notifications, or downstream integrations that rely on the completion endpoint being called (for example, if something outside RAGFlow listens for that POST), those won’t be triggered by a direct DB/ES update and would need to be handled separately. For most standard RAGFlow setups, you’re safe just updating the DB and ES directly, but for more complex workflows, double-check if you have any custom post-completion logic wired up elsewhere.
Reference

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}   

[voycey]

For others hitting this problem - I had to change the HOST_IP in settings.py to localhost as I am running in Docker. This is for the RAPTOR solution, I will try the same on GraphRAG to see if it solves the errors on that too!

HOST_IP = get_base_config(RAG_FLOW_SERVICE_NAME, {}).get("host", "localhost")

This really needs to be changed to be an environment variable so we can set in .env