feat: initial commit of the openlit translator

[91pavan]

• Add OpenLit translator, follows closely with what we did in the traceloop translator

Example trace dump of the resulting spans after running the example:

2999ad5b4013 :) select * from otel_traces\G;

SELECT *
FROM otel_traces

Query id: 95b25a41-4017-4384-aa7e-8ef05cfc02e4

Row 1:
──────
Timestamp:          2025-12-09 10:55:14.151148000
TraceId:            5eed9f9fea5522d88d8395446bb6ea9a
SpanId:             a88c55195a21c30b
ParentSpanId:       e6928e260ddd73b0
TraceState:
SpanName:           POST
SpanKind:           Client
ServiceName:        default
ResourceAttributes: {'deployment.environment':'default','service.name':'default','telemetry.sdk.language':'python','telemetry.sdk.name':'openlit','telemetry.sdk.version':'1.39.0'}
ScopeName:          opentelemetry.instrumentation.httpx
ScopeVersion:       0.60b0
SpanAttributes:     {'http.method':'POST','http.status_code':'200','http.url':'https://api.openai.com/v1/chat/completions'}
Duration:           2103346000 -- 2.10 billion
StatusCode:         Unset
StatusMessage:
Events.Timestamp:   []
Events.Name:        []
Events.Attributes:  []
Links.TraceId:      []
Links.SpanId:       []
Links.TraceState:   []
Links.Attributes:   []

Row 2:
──────
Timestamp:          2025-12-09 10:55:14.130954000
TraceId:            5eed9f9fea5522d88d8395446bb6ea9a
SpanId:             e6928e260ddd73b0
ParentSpanId:
TraceState:
SpanName:           genai.chat
SpanKind:           Client
ServiceName:        default
ResourceAttributes: {'deployment.environment':'default','service.name':'default','telemetry.sdk.language':'python','telemetry.sdk.name':'openlit','telemetry.sdk.version':'1.39.0'}
ScopeName:          opentelemetry.util.genai.handler
ScopeVersion:       0.1.4
SpanAttributes:     {'_openlit_processed':'true','deployment.environment':'default','gen_ai.client.token.usage':'124','gen_ai.evaluation.sampled':'true','gen_ai.input.messages':'[{"role": "user", "parts": [{"type": "text", "content": "user: What is LLM Observability?"}]}]','gen_ai.operation.name':'chat','gen_ai.output.messages':'[{"role": "assistant", "parts": [{"type": "text", "content": "LLM Observability stands for Logs, Metrics, and Traces observability. It refers to the practice of monitoring and measuring the performance and behavior of software systems by collecting and analyzing logs, metrics, and traces. Logs provide detailed records of events and activities within a system, metrics track key performance indicators and statistics, while traces enable understanding of the flow of requests and interactions between different components of a system. By leveraging LLM observability, organizations can gain insights into the health and performance of their systems, troubleshoot issues, and optimize resource utilization."}], "finish_reason": "stop"}]','gen_ai.output.type':'text','gen_ai.request.frequency_penalty':'0','gen_ai.request.is_stream':'false','gen_ai.request.max_tokens':'-1','gen_ai.request.model':'gpt-3.5-turbo','gen_ai.request.presence_penalty':'0','gen_ai.request.seed':'','gen_ai.request.stop_sequences':'[]','gen_ai.request.temperature':'1','gen_ai.request.top_p':'1','gen_ai.request.user':'','gen_ai.response.finish_reasons':'["stop"]','gen_ai.response.id':'chatcmpl-Ckpf9lY6sTPKKHqxKYj395bb355ym','gen_ai.response.model':'gpt-3.5-turbo-0125','gen_ai.sdk.version':'1.109.1','gen_ai.server.time_per_output_token':'0','gen_ai.server.time_to_first_token':'2.1309502124786377','gen_ai.service.tier':'default','gen_ai.system':'openai','gen_ai.usage.cost':'0.000172','gen_ai.usage.input_tokens':'14','gen_ai.usage.output_tokens':'110','server.address':'api.openai.com','server.port':'443','service.name':'default','telemetry.sdk.name':'openlit'}
Duration:           2131456000 -- 2.13 billion
StatusCode:         Ok
StatusMessage:
Events.Timestamp:   ['2025-12-09 10:55:16.262128000','2025-12-09 10:55:16.262137000']
Events.Name:        ['gen_ai.content.prompt','gen_ai.content.completion']
Events.Attributes:  [{'gen_ai.prompt':'user: What is LLM Observability?'},{'gen_ai.completion':'LLM Observability stands for Logs, Metrics, and Traces observability. It refers to the practice of monitoring and measuring the performance and behavior of software systems by collecting and analyzing logs, metrics, and traces. Logs provide detailed records of events and activities within a system, metrics track key performance indicators and statistics, while traces enable understanding of the flow of requests and interactions between different components of a system. By leveraging LLM observability, organizations can gain insights into the health and performance of their systems, troubleshoot issues, and optimize resource utilization.'}]
Links.TraceId:      []
Links.SpanId:       []
Links.TraceState:   []
Links.Attributes:   []

2 rows in set. Elapsed: 0.015 sec.

[pradystar]

Are we sure there will only be one completion response all the time? I mean would there be gen_ai.completion.1.content and so on?

[91pavan]

Good point. Usually it's one, but there could be more. will update the code to handle it

[pradystar]

I am concerned if both translator packages are installed one will override the other. I am not sure if both translators will be installed at the same time, but if they are then this a problem.

[91pavan]

The underlying framework could either be traceloop or openlit but not both right? So, for a single app, I doubt both will be installed. We could exit or warn the user about the presence of the other translator package.

Thoughts?

[pradystar]

maybe I am overthinking, but the execution environment could have both the packages installed and might lead to unexpected results. Maybe a warning or handling the wrapping in a thread safe manner and using wraps from functools?

[pradystar]

nit: OpenLit_PROCESSOR?

[91pavan]

fixed

[pradystar]

some test files names have to be updated -> replace traceloop with openlit?, example test_nested_traceloop_reconstruction.py