Cosortium acts as a proxy for LLM calls.
With 1 line of code change (API URL), Consortium lets you make any LLM powered app
- Smarter & Cheaper: via Ensemble LLM topologies, similar to OpenRouter's Fusion API. On some tasks like deep-research, code review, and Legal you can get Fable-level performance for 1-6% the price by fusing the responses of multiple cheaper models
- Reliable: Durable DAG execution improves the success rates of long/complex workflows by an order of magnitude. With result parsing, automatic retries, retry policies.
- Auditable: Full audit trails of every request, response, attempt, prompt, etc.
- Optimizable: Built-in wiring for industry-standard and custom benchmarks, with autoresearch/GEPA/MIPROv2-style optimization techniques.
This is not a replacement for n8n or similar. Consortium focuses on Ensemble LLM topologies, changing the way your LLM does its reasoning and forms an answer to support multiple topologies that fit different use cases.
Some use cases may need LLM-as-Judge, some may need Adverserial debate, some may need simple fusion (informed capitan), and some may need self-consistency. With consortium, thats a 1 click change. Your application code remains unchanged.
We have finite brain power.
The power you dedicate to thinking about what hotkey (/combination of) to press for niche tasks, is not, per say, wasted. But it definitely isn't going towards thinking about what to write, the objective.
Maybe useful for less information-dense writing.
>What are the use cases for this?
Mainly, version control, but also, this enables any usage that expects File System (files) as an input.
It is much easier to handle files when ingesting and processing data.
Below are some usecases i am thinking about (I will soon develop some of this):
1. SQL Formatter
2. A.I powered documentation
3. Technical debt management w/ dependecy management techniques
>If I make changes to the repo files can they be synces back to the Metabase database?
Not yet, but easily doable via current code architecture.
I suggest the following if the user hits 5000 submissions:
Define the 5000 as a soft limit.
Define something like 10,000 or "100 days after reaching the soft limit" as a hard limit.
When a user hits soft limit, you keep storing their submissions, but you don't let them see it unless they upgrade the subscription/buy more resources (Mailchimp's Blocks pricing strategy seems sensible to me). When a submission is received after hitting soft limit, you send your user an email telling them that they received a form submission, but they cant view it till they upgrade. (You keep notifying them). You can aggregate those notifications daily/weekly after the first few to reduce costs.
When the hard limit is reached, you stop storing those entries. And send the user 1 email telling him that you changed behaviour.
Here is my story. The 10X Broken-Engineer. I broke 10 bones in a Road Traffic Accident Oct 2020.
Broke: Femur, Tibia, fibula, Humerus, Ulna, Radius, and 4 ribs. And then, re-broke Tibia again.
I don't claim that i handled it well, nor that i now know how, but i can surely tell you what i did right and what i did wrong.
- Stay self-vigilant, especially on things related to your psyche. Staying bedridden or indoors for a long time will surely affect your hormone levels and mood. Remind yourself of this on your first tantrum.
- Start physical therapy AS SOON AS your doctor clears you to. Best choice i made. I know people whose joints are "limited" in movement range permanently, just because they skipped physio. Pain is temporary, joint range limitation is permanent.
- If you are getting a cast, it is so worth it to buy a "cast cover". This will make taking showers much easier.
- STOP SMOKING (If you are). I took it as the usual "stop smoking" that felt like an obligatory advice doctors had to give. Bone healing was basically non-existent for MONTHS. Quit smoking, started seeing remarkable progress nearly immediately.
- Ignore your pain. This is what i did best, and has probably contributed A TON to my recovery. I made sure to double check every "kind" of pain with my doc, to make sure nothing was worrying. As soon as the doc told me that it was normal, I started ignoring it. Think of it as background noise. No need to stop it. Just ignore it. Keep pushing yourself to the limit that your doc allows.
- Dont ignore too much pain. You may break a bone again. You are not bulletproof now. Even if you have Titanium plates and screws holding your bones together, you can still break them. Just take my word on this one .
- If you think bone placement/length/angle/feeling isn't right, dont wait for the bone to heal (cast removed) to check it. Go get a second opinion RN! The last thing you want is to have them break the bone and fixate it again.
- If you are going to have non-professional caretakers (friends/family), they need to know that your short temper and frustration is not towards them.
Golang integration with Clockwork, a tool (Chrome Extension) developed by the PHP community, for runtime insights.
Or as put in Clockwork's repo:
"lockwork gives you an insight into your application runtime - including request data, performance metrics, log entries, database queries, cache queries, redis commands, dispatched events, queued jobs, rendered views and more - for HTTP requests, commands, queue jobs and tests."
I just wish more components exist.
I struggled to decide where my async data pipeline ends.
There should be a generic component (something like "Cloud") that i can use when 3rd parties are actioned (i.e. push notifications, emails, log, api trigger).
- Smarter & Cheaper: via Ensemble LLM topologies, similar to OpenRouter's Fusion API. On some tasks like deep-research, code review, and Legal you can get Fable-level performance for 1-6% the price by fusing the responses of multiple cheaper models
- Reliable: Durable DAG execution improves the success rates of long/complex workflows by an order of magnitude. With result parsing, automatic retries, retry policies.
- Auditable: Full audit trails of every request, response, attempt, prompt, etc.
- Optimizable: Built-in wiring for industry-standard and custom benchmarks, with autoresearch/GEPA/MIPROv2-style optimization techniques.
This is not a replacement for n8n or similar. Consortium focuses on Ensemble LLM topologies, changing the way your LLM does its reasoning and forms an answer to support multiple topologies that fit different use cases.
Some use cases may need LLM-as-Judge, some may need Adverserial debate, some may need simple fusion (informed capitan), and some may need self-consistency. With consortium, thats a 1 click change. Your application code remains unchanged.