What's Changed
1. UI Improvements by ishaan-jaff in https://github.com/BerriAI/litellm/pull/1575
New Endpoints - [Spend Tracking]
- `/spend/keys` - View all keys created, ordered by spend
- `/spend/logs` - View all spend logs, if request_id is provided, only logs for that request_id will be returned
Example response for `/spend/logs`
json
[{"request_id":"chatcmpl-6dcb2540-d3d7-4e49-bb27-291f863f112e","call_type":"acompletion","api_key":"51**","spend":7.25e-05,"startTime":"2024-01-18T18:07:54.071000Z","endTime":"2024-01-18T18:07:55.510000Z","model":"cohere.command-text-v14","user":"litellm-is-awesome-user","modelParameters":{},"messages":[{"role":"user","content":"what llm are you-444"}],"response":"{\"id\":\"chatcmpl-6dcb2540-d3d7-4e49-bb27-291f863f112e\",\"choices\":[{\"finish_reason\":\"stop\",\"index\":0,\"message\":{\"content\":\" I am an AI chatbot, trained as an LLM by Cohere. Is there anything else you would like to know about LLMs? \",\"role\":\"assistant\"}}],\"created\":1705630075,\"model\":\"cohere.command-text-v14\",\"object\":\"chat.completion\",\"system_fingerprint\":null,\"usage\":{\"prompt_tokens\":7,\"completion_tokens\":31,\"total_tokens\":38}}","usage":{"total_tokens":38,"prompt_tokens":7,"completion_tokens":31},"metadata":{"headers":{"host":"0.0.0.0:8000","accept":"*/*","user-agent":"curl/7.88.1","content-type":"application/json","authorization":"Bearer sk-zNhplEZxJ-yOuZkqRvwO_g","content-length":"224"},"endpoint":"http://0.0.0.0:8000/chat/completions","deployment":"bedrock/cohere.command-text-v14","model_group":"BEDROCK_GROUP","user_api_key":"2b6e4ff4c14f67b6ccd7318948f0d66fe8724099224baeaf4733792a2048aeb3","caching_groups":null,"user_api_key_user_id":"ishaan2berri.ai","user_api_key_metadata":{}},"cache_hit":""}]
2. [Feat] Make Proxy Auth Exceptions OpenAI compatible by ishaan-jaff in https://github.com/BerriAI/litellm/pull/1576
🚨 If you use Custom Auth on the Proxy, It will actually raise an now https://docs.litellm.ai/docs/proxy/virtual_keys#custom-auth
Auth Errors Before PR - using OpenAI JS/Python
js
AuthenticationError: 401 status code (no body)
at APIError.generate (/Users/ishaanjaffer/Github/litellm/litellm/proxy/tests/node_modules/openai/error.js:47:20)
at OpenAI.makeStatusError (/Users/ishaanjaffer/Github/litellm/litellm/proxy/tests/node_modules/openai/core.js:255:33)
at OpenAI.makeRequest (/Users/ishaanjaffer/Github/litellm/litellm/proxy/tests/node_modules/openai/core.js:294:30)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async runOpenAI (/Users/ishaanjaffer/Github/litellm/litellm/proxy/tests/test_openai_js.js:12:22) {
status: 401,
headers: {
'content-length': '117',
'content-type': 'application/json',
date: 'Mon, 22 Jan 2024 21:21:14 GMT',
server: 'uvicorn'
},
error: undefined,
code: undefined,
param: undefined,
type: undefined
}
Auth Errors After PR - using OpenAI JS/Python
js
AuthenticationError: 401 Authentication Error: invalid user key - token does not exist
at APIError.generate (/Users/ishaanjaffer/Github/litellm/litellm/proxy/tests/node_modules/openai/error.js:47:20)
at OpenAI.makeStatusError (/Users/ishaanjaffer/Github/litellm/litellm/proxy/tests/node_modules/openai/core.js:255:33)
at OpenAI.makeRequest (/Users/ishaanjaffer/Github/litellm/litellm/proxy/tests/node_modules/openai/core.js:294:30)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async runOpenAI (/Users/ishaanjaffer/Github/litellm/litellm/proxy/tests/test_openai_js.js:12:22) {
status: 401,
headers: {
'content-length': '131',
'content-type': 'application/json',
date: 'Tue, 23 Jan 2024 00:26:52 GMT',
server: 'uvicorn'
},
error: {
message: 'Authentication Error: invalid user key - token does not exist',
type: 'auth_error',
param: 'None',
code: 401
},
code: 401,
param: 'None',
type: 'auth_error'
3. [Fix] Streaming - Use same `response_id` across chunks by ishaan-jaff in https://github.com/BerriAI/litellm/pull/1572
Streaming Responses (for non OpenAI/Azure) After this PR - `id` is the same across chunks
shell
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content='The', role='assistant'))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=' sky', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=',', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=' a', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=' canvas', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=' of', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=' blue', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=' and', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=' gold', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content=',', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2
ModelResponse(id='chatcmpl-8e5a8a5f-46ba-47d9-8858-c4d81aa4efc2', choices=[StreamingChoices(finish_reason=None, index=0, delta=Delta(content='\n', role=None))], created=1706043294, model='berri-benchmarking-Llama-2-70b-chat-hf-4', object='chat.completion.chunk', system_fingerprint=None, usage=Usage())
* [WIP] fix(utils.py): fix proxy streaming spend tracking by krrishdholakia in https://github.com/BerriAI/litellm/pull/1574
* feat(proxy/utils.py): Support keys with budget duration (resets spend on duration end) by krrishdholakia in https://github.com/BerriAI/litellm/pull/1570
**Full Changelog**: https://github.com/BerriAI/litellm/compare/v1.18.11...v1.18.12