Mukunda Rao Katta

May 28

Different agent frameworks log 'input_tokens' differently. Here's the fix.

#devchallenge #hermesagentchallenge #agents #python

400 from Anthropic API. Two consecutive user messages. Every time.

2 min read

Want to connect with Mukunda Rao Katta?

Create an account to connect with Mukunda Rao Katta. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

Hermes Agent Challenge Submission: Write About Hermes Agent

May 28

400 from Anthropic API. Two consecutive user messages. Every time.

#devchallenge #hermesagentchallenge #agents #python

Context window exceeded at turn 23. Here's how I track token usage without a tokenizer.

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

Context window exceeded at turn 23. Here's how I track token usage without a tokenizer.

#devchallenge #hermesagentchallenge #agents #python

My Hermes agent spent 3ドル before I noticed. Now it can't.

1 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

My Hermes agent spent 3ドル before I noticed. Now it can't.

#devchallenge #hermesagentchallenge #agents #python

My Hermes agent's stop condition was a 40-line if/elif chain. I replaced it with 3 lines.

1 comment

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

My Hermes agent's stop condition was a 40-line if/elif chain. I replaced it with 3 lines.

#devchallenge #hermesagentchallenge #agents #python

My agent kept hitting context limits. This one function fixed it.

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

My agent kept hitting context limits. This one function fixed it.

#devchallenge #hermesagentchallenge #agents #python

Your Hermes agent's audit log is leaking customer emails. Here's a 100-line lib that fixes that.

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

My agent kept forgetting what it was doing. A scratchpad fixed it.

May 25

Your Hermes agent's audit log is leaking customer emails. Here's a 100-line lib that fixes that.

#devchallenge #hermesagentchallenge #agents #security

1 reaction

2 comments

4 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

My agent kept forgetting what it was doing. A scratchpad fixed it.

#devchallenge #hermesagentchallenge #agents #python

I replaced 200 lines of ad-hoc state management in my Hermes agent with one object.

2 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

I replaced 200 lines of ad-hoc state management in my Hermes agent with one object.

#devchallenge #hermesagentchallenge #agents #python

The two-line Hermes agent logger I wish existed a month ago

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

The two-line Hermes agent logger I wish existed a month ago

#devchallenge #hermesagentchallenge #agents #observability

My Hermes agent called exec_shell. It shouldn't have been able to.

3 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

My Hermes agent called exec_shell. It shouldn't have been able to.

#devchallenge #hermesagentchallenge #agents #python

My Hermes agent ran 500 turns and cost 40ドル before I noticed. Now it can't.

2 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

My Hermes agent ran 500 turns and cost 40ドル before I noticed. Now it can't.

#devchallenge #hermesagentchallenge #agents #python

My Hermes agent loop blew the context window at turn 47. llm-context-trim fixed it.

2 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

My Hermes agent loop blew the context window at turn 47. llm-context-trim fixed it.

#devchallenge #hermesagentchallenge #agents #python

Building LLM message lists by hand is error-prone. There's a better way.

3 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

Building LLM message lists by hand is error-prone. There's a better way.

#devchallenge #hermesagentchallenge #agents #python

My Hermes agent kept returning JSON in a markdown code block. I kept writing the same regex. Then I stopped.

2 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

My Hermes agent kept returning JSON in a markdown code block. I kept writing the same regex. Then I stopped.

#devchallenge #hermesagentchallenge #agents #python

Claude's extended thinking is useful. Sending raw <thinking> tags to users is not.

2 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

Claude's extended thinking is useful. Sending raw <thinking> tags to users is not.

#devchallenge #hermesagentchallenge #agents #python

I changed one line in my Hermes system prompt. prompt-diff showed me exactly what.

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

I changed one line in my Hermes system prompt. prompt-diff showed me exactly what.

#devchallenge #hermesagentchallenge #agents #prompts

Your Hermes agent's prompt cache is 30% hit because of one timestamp. Here's how to find it.

3 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

Your Hermes agent's prompt cache is 30% hit because of one timestamp. Here's how to find it.

#devchallenge #hermesagentchallenge #agents #performance

My agent started lying. It took me two days to realize the system prompt had changed.

4 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

My agent started lying. It took me two days to realize the system prompt had changed.

#devchallenge #hermesagentchallenge #agents #python

I changed my Hermes agent's system prompt and used tool-call-diff to prove it actually helped

3 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

I changed my Hermes agent's system prompt and used tool-call-diff to prove it actually helped

#devchallenge #hermesagentchallenge #agents #testing

I couldn't reproduce the bug because I had no record of what tool calls my Hermes agent made.

4 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

I couldn't reproduce the bug because I had no record of what tool calls my Hermes agent made.

#devchallenge #hermesagentchallenge #agents #python

My agent called web_search 47 times in 10 seconds. One class stopped it.

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

My agent called web_search 47 times in 10 seconds. One class stopped it.

#devchallenge #hermesagentchallenge #agents #python

1 reaction

One Hermes tool call took 4.8 seconds. The others averaged 900ms. trace-anomaly found it.

2 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

One Hermes tool call took 4.8 seconds. The others averaged 900ms. trace-anomaly found it.

#devchallenge #hermesagentchallenge #agents #observability

My Hermes agent ran for 2 hours, hit a timeout at turn 89, and I had to start over. Now it checkpoints.

3 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

My Hermes agent ran for 2 hours, hit a timeout at turn 89, and I had to start over. Now it checkpoints.

#devchallenge #hermesagentchallenge #agents #python

My Hermes agent spent how much? trace-cost for JSONL audit logs

2 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

My Hermes agent spent how much? trace-cost for JSONL audit logs

#devchallenge #hermesagentchallenge #agents #observability

I had 3,000 agent trace events in JSONL. My manager wanted them in Excel. One line.

1 comment

3 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

I had 3,000 agent trace events in JSONL. My manager wanted them in Excel. One line.

#devchallenge #hermesagentchallenge #agents #python

5000 events, one worker, one bug: trace-filter for agent JSONL traces

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

5000 events, one worker, one bug: trace-filter for agent JSONL traces

#devchallenge #hermesagentchallenge #agents #observability

grep for your agent traces — substring, exact, and regex in one call

4 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

grep for your agent traces — substring, exact, and regex in one call

#devchallenge #hermesagentchallenge #agents #python

I ran three Hermes agents in parallel. trace-merge stitched them into one timeline I could read.

2 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

I ran three Hermes agents in parallel. trace-merge stitched them into one timeline I could read.

#devchallenge #hermesagentchallenge #agents #observability

I logged 300 Hermes runs to one file. trace-session-split cut it into 300.

5 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

I logged 300 Hermes runs to one file. trace-session-split cut it into 300.

#devchallenge #hermesagentchallenge #agents #observability

p95 latency was 1.68 seconds. I didn't know until I built trace-stats.

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

p95 latency was 1.68 seconds. I didn't know until I built trace-stats.

#devchallenge #hermesagentchallenge #agents #observability

After every Hermes agent run, I print a 10-line summary. Now I know exactly what happened.

3 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

After every Hermes agent run, I print a 10-line summary. Now I know exactly what happened.

#devchallenge #hermesagentchallenge #agents #python

Pipe your Hermes Agent audit log into Datadog in 5 lines (no OTel SDK required)

2 min read

Hermes Agent Challenge Submission: Write About Hermes Agent

May 25

Pipe your Hermes Agent audit log into Datadog in 5 lines (no OTel SDK required)

#devchallenge #hermesagentchallenge #observability #agents

I had 800 lines of Hermes agent audit log. trace-tree turned it into a tree I could read.

5 min read

Hermes Agent Challenge Submission: Build With Hermes Agent

May 25

I had 800 lines of Hermes agent audit log. trace-tree turned it into a tree I could read.

#devchallenge #hermesagentchallenge #agents #observability

I shipped 50 agent infrastructure libraries. Here is what I learned.

5 min read

May 25

I shipped 50 agent infrastructure libraries. Here is what I learned.

prompt-shield: a tiny, zero-dep prompt-injection detector you can drop in front of any agent

7 min read

May 25

prompt-shield: a tiny, zero-dep prompt-injection detector you can drop in front of any agent

A six-concern production harness for Nemotron agents on Crusoe Managed Inference

5 min read

May 25

A six-concern production harness for Nemotron agents on Crusoe Managed Inference

agenttap: see exactly what your LLM SDK sent to the wire, with API keys scrubbed

4 min read

May 25

agenttap: see exactly what your LLM SDK sent to the wire, with API keys scrubbed

llmfleet: pool many agents' turns into one Batch API call and save 50 percent

4 min read

May 25

llmfleet: pool many agents' turns into one Batch API call and save 50 percent

cachebench: stop finding out about prompt-cache regressions from the invoice

4 min read

May 25

cachebench: stop finding out about prompt-cache regressions from the invoice

I shipped eight agent-stack repos in eight hours. Here's what made it possible.

4 min read

May 25

I shipped eight agent-stack repos in eight hours. Here's what made it possible.

I needed a stable cache key for LLM requests. The hard part was the input list order.

4 min read

May 25

I needed a stable cache key for LLM requests. The hard part was the input list order.

#hermeschallenge #ai #llm #rust

My LLM provider went down for 11 minutes. My code spent 4 of them in connect timeouts.

4 min read

May 25

My LLM provider went down for 11 minutes. My code spent 4 of them in connect timeouts.

#hermeschallenge #ai #llm #rust

My RAG system slowly got worse for three months and nobody noticed.

4 min read

May 25

My RAG system slowly got worse for three months and nobody noticed.

#hermeschallenge #ai #rag #python

I rewrote my Bedrock client three times. Then I extracted the boring parts into one library.

4 min read

May 25

I rewrote my Bedrock client three times. Then I extracted the boring parts into one library.

#hermeschallenge #aws #bedrock #python

I Shipped 50 Agent-Stack Libraries. Here Is How They Fit Together.

4 min read

May 25

I Shipped 50 Agent-Stack Libraries. Here Is How They Fit Together.

prompt-eval-rubric: Score Your Agent's Outputs Without Paying for Another LLM Call

6 min read

May 25

prompt-eval-rubric: Score Your Agent's Outputs Without Paying for Another LLM Call

agent-fn-registry: Register Your Agent Tools With Schema, Side Effects, and Defaults in One Place

5 min read

May 25

agent-fn-registry: Register Your Agent Tools With Schema, Side Effects, and Defaults in One Place

agent-replay-trace: Load and Step Through Agent Traces for Debugging

5 min read

May 25

agent-replay-trace: Load and Step Through Agent Traces for Debugging

agent-shadow-mode: Run Your Agent in Record-Not-Execute Mode Before You Trust It

5 min read

May 25

agent-shadow-mode: Run Your Agent in Record-Not-Execute Mode Before You Trust It

llm-pii-redact: Remove PII From Prompts Before They Reach the Model

5 min read

May 25

llm-pii-redact: Remove PII From Prompts Before They Reach the Model

Why I Built 50 Small Libraries Instead of One Big Agent Framework

6 min read

May 25

Why I Built 50 Small Libraries Instead of One Big Agent Framework

Production Python Agent Checklist: 15 Things Before You Deploy

1 comment

6 min read

May 25

Production Python Agent Checklist: 15 Things Before You Deploy

When and How to Use the Anthropic Batch API in Your Agent

6 min read

May 25

When and How to Use the Anthropic Batch API in Your Agent

What I Learned Shipping 50 Agent-Stack Libraries in One Sprint

6 min read

May 25

What I Learned Shipping 50 Agent-Stack Libraries in One Sprint

My agent called search() with `{"query": null}`. The error message I returned saved the next 30 minutes.

6 min read

May 25

My agent called search() with `{"query": null}`. The error message I returned saved the next 30 minutes.

How to Test LLM Agents Without Calling the Real API

4 min read

May 25

How to Test LLM Agents Without Calling the Real API

Make Your Agent's API Calls Idempotent Before You Need To

5 min read

May 25

Make Your Agent's API Calls Idempotent Before You Need To

Four Patterns for Multi-Agent Python Systems That Actually Work

5 min read

May 25

Four Patterns for Multi-Agent Python Systems That Actually Work

Three Error Recovery Patterns for LLM Agent Tool Failures

5 min read

May 25

Three Error Recovery Patterns for LLM Agent Tool Failures