DOC-797 | Natural language to AQL service (txt2aql / AQLizer) #794
Conversation
|
Deploy Preview Available Via |
…o longer needed, add aqlizer mode
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
There was a problem hiding this comment.
Thank you for rasing this PR @nerpaula, I have added a few comments for consideration.
site/content/3.12/data-science/graphrag/services/natural-language-to-aql.md
Outdated
Show resolved
Hide resolved
| --header 'Content-Type: application/json' \ | ||
| --data '{ | ||
| "env": { | ||
| "username": "<your-username>", |
There was a problem hiding this comment.
Note: username is currently still required for startup, but we are working on removing this requirement (following the same approach in the importer and retriever).
There was a problem hiding this comment.
I have removed it from now but will double check to see when the upstream PRs get merged.
| {{< /info >}} | ||
|
|
||
| ### Configuration | ||
| ## Installation and configuration |
There was a problem hiding this comment.
This section switches between runtime environment variables and startup parameters, which could cause some confusion for the reader. It would probably be easier to read when consistently referring to the startup parameters, which are required to deploy the service - what do you think?
There was a problem hiding this comment.
Good point! Reading it all over again made me realize it was indeed unclear and likely causing confusion. I have entirely restructured this section and consistently referred to startup parameters - users didn't need to worry about managing env variables. The steps should now be more intuitive and just enough information to succeed at each step.
|
|
||
| - ArangoDB instance | ||
| - OpenAI API key (if using OpenAI as provider) | ||
| - Triton URL and model name (if using Triton as provider) |
There was a problem hiding this comment.
Has it been decided to completely remove the mention of Triton? It is technically supported, but possibly won't be used a lot.
There was a problem hiding this comment.
| ```bash | ||
| curl --request POST \ | ||
| --url https://<ExternalEndpoint>:8529/ai/v1/graphrag \ | ||
| --header 'Authorization: Bearer <your-bearer-token>' \ |
There was a problem hiding this comment.
Suggestion: Add a reference on how this token can be obtained.
There was a problem hiding this comment.
Added a separate section about this.
| --header 'Content-Type: application/json' \ | ||
| --data '{ | ||
| "input_text": "What are the advantages of graph databases?", | ||
| "mode": "aqlizer" |
There was a problem hiding this comment.
Some clarification is needed here. The process_text_stream works with and without "mode": "aqlizer". When using the default mode, the above documentation is correct.
However, with "mode": "aqlizer" the endpoint has a completely different behavior. It converts natural language questions into AQL queries by streaming responses from an LLM that has knowledge of the database schema (reference PR).
There was a problem hiding this comment.
I tried to clarify this - split it into sections (default mode and aqlizer mode) and added examples. Hope it's clear enough now.
Description
Natural language to AQL service (txt2aql / AQLizer) documentation.
Note
Rewrite and expand service docs with deployment/config, streaming (AQLizer) support, updated REST endpoints and output formats, OpenAI/OpenRouter config, and improved guidance; remove gRPC/Triton specifics.
graphragNatural Language to AQL serviceprocess_textandtranslate_query.POST /ai/v1/graphrag, retrieveserviceId, verify status, and health check.envparams; remove Triton-specific details.POST /v1/process_textand newPOST /v1/process_text_streamwith default and AQLizer modes (streaming AQL generation).POST /v1/translate_querywith output formats (NL,AQL,JSON) and concise examples.Written by Cursor Bugbot for commit 238ba4a. This will update automatically on new commits. Configure here.