 |
| https://www.pexels.com/photo/a-man-with-a-mustache-in-white-long-sleeves-7568794/ |
In the previous few blogs, we succinctly look at the different AutoGen orchestrators. In this blog, we compare them, essentially looking that number of turns needed to complete a simple task, and the number of token usages.
The code is in this repo. Below are the results of sample successful runs
SelectorGroupChat
```json
{
"account_id": "789",
"saving_balance": 30000.3,
"investment_balance": 300000.3,
"total_balance": 330000.6
}
```
TERMINATE
Stop reason: Text 'TERMINATE' mentioned
Agent Name Prompt Tokens Completion Tokens
-------------------- --------------- -------------------
saving_account_agent 99 32
account_agent 102 12
saving_account_agent 146 19
customer_agent 254 36
investment_agent 202 20
customer_agent 308 51Round Robin Orchestrator
It went a few turns before landed with a solution. We can see that the number of tokens in prompts grow over time.
```json
{
"account_id": "123",
"saving_balance": 10000.1,
"investment_balance": 100000.1,
"total_balance": 110000.2
}
```
TERMINATE
Stop reason: Text 'TERMINATE' mentioned
Agent Name Prompt Tokens Completion Tokens
-------------------- --------------- -------------------
customer_agent 195 9
account_agent 79 12
saving_account_agent 123 19
investment_agent 136 39
customer_agent 276 9
account_agent 178 30
saving_account_agent 230 31
investment_agent 243 20
customer_agent 349 51
Swarm
When we are not careful with Swarm, We can end up like this. many turns and a lot of token used.
```json
{
"account_id": "456",
"saving_balance": 20000.2,
"investment_balance": 200000.2,
"total_balance": 220000.4
}
```
TERMINATE
Stop reason: Text 'TERMINATE' mentioned
Agent Name Prompt Tokens Completion Tokens
-------------------- --------------- -------------------
customer_agent 243 45
account_agent 99 43
customer_agent 378 12
account_agent 193 32
account_agent 231 12
customer_agent 474 14
saving_account_agent 283 19
saving_account_agent 321 33
saving_account_agent 361 28
investment_agent 355 20
investment_agent 394 36
investment_agent 436 19
investment_agent 461 12
investment_agent 479 12
investment_agent 497 17
investment_agent 520 23
investment_agent 549 15
investment_agent 570 16
investment_agent 592 17
investment_agent 615 12
investment_agent 633 13
investment_agent 652 32
investment_agent 690 55
investment_agent 781 51
investment_agent 838 44
investment_agent 888 52
investment_agent 965 20
investment_agent 1004 20
investment_agent 1043 36
investment_agent 1085 15
investment_agent 1106 19
investment_agent 1131 16
investment_agent 1153 12
investment_agent 1171 17
investment_agent 1194 12
investment_agent 1212 17
investment_agent 1235 50
customer_agent 1316 52
Comments
Post a Comment