Post

Ramya Chinnadurai 🚀
Ramya Chinnadurai 🚀@code_rams·
Just updated my 5-agent setup based on @AlexFinn's breakdown: - Chiti (coordinator): Kimi K2.5 - Vasi (dev): MiniMax 2.1 - Sana (marketing): Kimi K2.5 - Nila (content): Kimi K2.5 - Boran (finance): DeepSeek V3 - Heartbeat: Haiku Previously using shared model = context overflow disaster. Now each agent has specialized model for their task. Testing with $5.63 openrouter credits left. Will report back. h/t @AlexFinn for the model breakdown 🙏 x.com/AlexFinn/statu…
Ramya Chinnadurai 🚀 tweet media
Ramya Chinnadurai 🚀@code_rams

Update on the $10/month AI bot experiment - reality hit hard! Here is how it went down. I'm running a multi-agent setup: - Chiti (coordinator) - Vasi (dev) - Sana (marketing) - Nila (content) - Boran (finance) Yesterday Sana did SEO research for @TweetsMashApp marketing and updated in Notion. Everything smooth. Beautiful. Then I asked Chiti to automate invoices for monthly taxes. BOOM. Continuous "Context overflow" issue. All agents crashed. Massive blackout. Gemini CLI OAuth? Soft-banned within hours. OpenAI Codex OAuth? Soft-ban Jail. Free models? "Model not allowed" errors. The fallback wasn't switching automatically. Had to manually dig into config.json and debug. Switched to openrouter/moonshotai/kimi-k2.5. Bot's alive again but experiencing latency issues. Current status: - OpenRouter credit: $5.63 left of $10 - Total spent debugging: ~$4 - Primary model: kimi-k2.5 (working but slow) - OAuth: disabled for now (soft-ban risk) The hard truth from @ImNotTheWolf - "It can't be cracked cheaply yet. Clawdbot basically NEEDS Opus 4.5 or GPT-5.2-Codex for reliability. Need to wait for another 6 months to make it less cost". But I'm not giving up. Let me try fixing the context management first. Chiti coordinating 5 subagents = context exploding to 50K+ tokens. OAuth won't save you if your agents keep bloating the context window. What I'm trying next: 1. Qdrant vector memory - externalize context so prompts stay small, bot queries on-demand instead of carrying 50K+ tokens (h/t @mej26vPIJx22495 for this suggestion in Tamil 🙏) 2. QMD local search - lightweight on-device doc search to reduce API calls 3. Heartbeat + Auto-Doctor - batch non-urgent tasks, monitor context size, auto-restart before overflow (h/t @ImNotTheWolf for sharing these scripts 🙏) 4. One model per subagent - isolate context per agent instead of shared bloating. Maybe @ImNotTheWolf is right. Maybe cheap models can't handle complex multi-agent coordination. Truth is, output from Opus 4.5 is extraordinary. Nothing beats it right now. But we can't wait until it's affordable for everyone. Until then, let's figure out how to power our bots with less cost. What if the problem isn't the model intelligence - it's how we manage memory? Will find out. Experimenting with $5.63 left. Will share the Qdrant + QMD setup once I test it. Any recommendations/suggestions on this is appreciated.

English
36
13
357
60.2K
Steen
Steen@RcSteen76·
@code_rams @AlexFinn You will still burn through your credits in no time flat. I went through $50 with kimi 2.5 in 1 day.
English
0
0
0
735
ChrisV.btc⚡
ChrisV.btc⚡@ChrisVolkernick·
@code_rams @AlexFinn Was literally just watching this video lmk how it goes 👀 I'm in token timeout until Saturday 🥴
English
0
0
0
1.1K
JayShugito
JayShugito@JayShugito·
@code_rams @AlexFinn my AI clued me in that you can use OpenRouter for this with just 1 api key instead of several. free package too.
English
0
0
0
123
Tj
Tj@thiru1209·
@code_rams @AlexFinn Interesting names - I m from shaktiman era - you gave me a thought to name agents like gangadhar, shaktiman, geetha viswas, tamraj kilwish 😀
English
0
0
0
812
Rohan
Rohan@rohan_x2·
@code_rams @AlexFinn How do you set-up dynamic routing of models? Is it done automatically once you config and add them in fallbacks?
English
0
0
0
486
Jim Panehal
Jim Panehal@JimPanehal·
@code_rams @AlexFinn It works great! I’ve set it up and using kimi 2.5 for free at the moment thru NVidia for all of them. Till NVidia starts charging. Then back to Alex setup.
English
0
0
0
330
Aman Rawat
Aman Rawat@amanrawatamg·
@code_rams @AlexFinn That context blackout is a huge pain once you get past a few turns. We’re testing a setup where we summarize the key points before handing off to the next agent so the context stays clean. Are you trying a similar 'summary handoff' yet or just letting the history pile up?
English
0
0
0
308
thindery
thindery@thindery·
@code_rams @AlexFinn I'm using @ollama $20 month plan with Kimi k2. Working well for now. Nice not worrying about cost thresholds. The usage limits are very generous with resets every 3 hours and 5 days
English
0
0
0
186
SlickNick
SlickNick@2sliccNick·
@code_rams @AlexFinn What took you guys so long to figure it out !? Claude recommended it to me on day 1 , I just had to ask about the advantages to running locally
English
0
0
0
181
Chris
Chris@chris_ships·
@code_rams @AlexFinn you should make an another agent to monitor costs and assign models dynamically.
English
0
0
0
34
E__Strobel
E__Strobel@E__Strobel·
@code_rams @AlexFinn So, you’re getting all that through OpenRouter? I’d heard there were problems configuring OpenClaw to use models off OpenRouter. Would love to hear more.
English
0
0
0
145
Social Monkey
Social Monkey@socialmonkeyai·
@code_rams @AlexFinn Interesting setup! I find the choice of models for each agent really shapes the overall output. I've been experimenting with a similar approach, focusing on the specific strengths of each model for the task at hand.
English
0
0
0
64
Aman Rawat
Aman Rawat@amanrawatamg·
@code_rams @AlexFinn Splitting the models is a heavy lift but the only way to beat context limits. Are you planning to implement a summary handoff between Chiti and the specialists? Otherwise that coordinator context will rot fast even with Kimi's window.
English
0
0
0
295
Wesman
Wesman@thewesguy·
@code_rams @AlexFinn Thanks for sharing. What do you mean exactly with the shared model causing issues?
English
0
0
0
262
Nitesh Rijal
Nitesh Rijal@opynrijal·
@code_rams @AlexFinn Nice. Is there a config file somewhere to look at and replicate this setup? Are you burning through $10/day or has that gone up?
English
0
0
0
6
Josh Ridener
Josh Ridener@joshridener85·
@code_rams @AlexFinn How do you set these up with out the agent stalling out. Every time I’ve tried to add multiple models I lose the gateway and have get it back through terminal and back at square 1
English
0
0
0
101
Paylaş