Skip navigation
Applied AI engineer's Night working at multiple projects for his company -
0:00 / 0:36

Applied AI engineer's Night working at multiple projects for his company -

Siddharth

@Pseudo_Sid266.4K followers

Visit on X
181 likesDownloadOriginal post

11,978 views

Applied AI engineer's Night working at multiple projects for his company - > Added Langfuse for observability and tracing of the voice agent pipeline. This tool never fails to amaze me. > Langfuse actually captured all the voice agent logs - per turn - that too real time - and that too in a nested concatenation ( made user_id the trace_id to pull this off). Absolute gold !! > Changed a deployment with some changes in db as well. Did some additions in fly to sustain the new backend. > Had a meet with design engineer and literally pulled off a random hour long discord session with my CTO on sunday night ( yesterday ) to discuss about potential new methods for long running agents. > Made some registries in the micro-kernel inspired architecture for memory integration and made ports and guards for agents to be able to access the memory tools ( hexagonal arch ) > Was watching portugal vs spain on the other screen with friends only to see my goat cry at the end :\

25 Comments

Comments from the original post on X (Twitter)

  • Amit's profile picture

    @actual_amit

    Thanks for reminding me of Langfuse.

    1 like

  • Siddharth's profile picture

    @Pseudo_Sid26

    Thanks but one sec, what do you do at merc ?

  • Amit's profile picture

    @actual_amit

    I joined 2 months ago. Role is AI Test Engineer, mostly working on evaluation, monitoring, and testing of AI things, but also other AI-related tasks.

  • Gaurav Singh's profile picture

    @_iamgsr_

    Love this! Langfuse is a game-changer for observability.

    1 like

  • Taqui's profile picture

    @md_taqui_imam

    Super productive bro

    1 like

  • Avid's profile picture

    @Av1dlive

    This is so awesome! Love it.

    1 like

  • Siddharth's profile picture

    @Pseudo_Sid26

    Thanks mate

  • Roshan's profile picture

    @itsroshanharry

    Alright today I'm gonna look in to langfuse. Awesome content mate!

    1 like

  • Siddharth's profile picture

    @Pseudo_Sid26

    Thanks

    1 like

  • Rishi's profile picture

    @Rishi_Sharma___

    Bruh how do you even know these many technologies ? U learn them or based on ai ? I'm confused these days how are ppl doing such things

    1 like

  • Siddharth's profile picture

    @Pseudo_Sid26

    Most of them come with exp as you build more n more. Even I don’t know everything. I keep learning

  • Rishi's profile picture

    @Rishi_Sharma___

    Impressive

  • AniketXD's profile picture

    @aniketrawat00

    We also use langfuse and it is honestly kind of godgiven provided we don't have to hardcode the prompts and redeploy everytime we make the slightest change + all the observability features and u can make a self improving pipeline around it also

    1 like

  • Siddharth's profile picture

    @Pseudo_Sid26

    +1, goated tool

    1 like

  • Roboki's profile picture

    @ItsRoboki

    W keep working 🔥

    1 like

  • Siddharth's profile picture

    @Pseudo_Sid26

    Yessir alwayssss

    1 like

  • Sunil Maurya's profile picture

    @sunilmaurya_25

    Insightful! Today I learned about Langfuse. Please share more topics like this about voice agents for non-technical people like me 😅

    1 like

  • Siddharth's profile picture

    @Pseudo_Sid26

    Suree

  • Adish Jain's profile picture

    @adishjain333

    Langfuse is great for this — it shows you what the voice agent did after each call. The complement is pre-deploy: the failures that hurt voice agents (skipped a verification step, wrong tool, right answer via a broken path) are cheaper to catch before shipping by testing the trajectory. Observability + behavioral testing is the combo. Building the second half OSS:

    1 like

  • Siddharth's profile picture

    @Pseudo_Sid26

    Will check this out