[ 
https://issues.apache.org/jira/browse/HDDS-15570?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Arafat Khan reassigned HDDS-15570:
----------------------------------

    Assignee: Arafat Khan

> [Recon-Chatbot] Call Recon methods directly instead of using HTTP-calls
> -----------------------------------------------------------------------
>
>                 Key: HDDS-15570
>                 URL: https://issues.apache.org/jira/browse/HDDS-15570
>             Project: Apache Ozone
>          Issue Type: Sub-task
>          Components: Ozone Recon
>            Reporter: Arafat Khan
>            Assignee: Arafat Khan
>            Priority: Major
>
> h2. Problem
> The Recon chatbot runs inside the same JVM as Recon. Making HTTP calls to 
> Recon APIs from there adds extra work: building URLs, parsing responses, and 
> duplicating logic that Recon already has in code.
> h2. Proposal
> Since the chatbot and Recon share the same process, the chatbot should call 
> Recon methods directly when it needs cluster data the same way Recon’s own 
> code would instead of going out over HTTP and back in.
> Use LangChain’s native tool-calling support so the LLM picks from a fixed 
> list of allowed tools (for example “get datanodes”, “list keys in a bucket”). 
> The chatbot then maps each tool name to the right Recon method and runs it 
> directly.
> h2. Why this helps
>  * Faster and simpler - no HTTP round trip inside the same server
>  * Fewer moving parts - less parsing, less error handling for network failures
>  * Clearer contract - the LLM chooses a tool; Java runs the matching Recon 
> logic
>  * Fits LangChain’s model - tool definitions, tool selection, and tool 
> execution in one flow
> h2. What stays the same
>  * Users still ask questions in natural language
>  * Safety rules still apply (allowed tools only, record limits, scoped 
> queries where needed)
>  * The public Recon REST API for the UI is unchanged
> h2. Done when
>  * Chatbot fetches Recon data via direct method calls, not HTTP to itself
>  * LLM tool selection uses LangChain native tools end to end
>  * Existing safety and limits still work



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to