Getting a 404 error from langfuse in a spring ai + kotlin app

[laskaridis]

Describe your question

Getting the following error from langfuse:

[angfuse.com/...] [                                                 ] i.o.exporter.internal.http.HttpExporter  : Failed to export logs. Server responded with HTTP status code 404. Error message: Unable to parse response body, HTTP status message:

Followed the setup instructions found here: https://langfuse.com/integrations/frameworks/spring-ai

build.gradle.kts:

dependencies {
  implementation("org.springframework.boot:spring-boot-starter-web")
  implementation("org.springframework.boot:spring-boot-starter-actuator")
  implementation("org.springframework.ai:spring-ai-starter-model-openai")
  implementation("io.opentelemetry.instrumentation:opentelemetry-spring-boot-starter")
  implementation("io.opentelemetry:opentelemetry-exporter-otlp")
  implementation("io.micrometer:micrometer-tracing-bridge-otel")
  ...
 } 
 ependencyManagement {
  imports {
    mavenBom("org.springframework.ai:spring-ai-bom:${property("springAiVersion")}")
    mavenBom("io.opentelemetry.instrumentation:opentelemetry-instrumentation-bom:${property("otelInstrumentationVersion")}")
  }
}

application.properties:

spring.application.name=llm-playground-service
spring.ai.openai.api-key=<my key>
spring.ai.chat.observations.log-prompt=true
spring.ai.chat.observations.log-completion=true
management.tracing.sampling.probability=1.0
management.observations.annotations.enabled=true
...

Using the following ENV vars:

OTEL_EXPORTER_OTLP_ENDPOINT=https://cloud.langfuse.com/api/public/otel/v1/traces
OTEL_EXPORTER_OTLP_HEADERS=Authorization=Basic <my base64 encoded public key>

Also tried the following with no success:

OTEL_EXPORTER_OTLP_ENDPOINT=https://cloud.langfuse.com/api/public/otel

I'm creating my authentication key from the CLI as follows:

$ echo -n <my public key> | base64

Can you please help? I'm trying to eval the platform for a project and this is time sensitive. Thanks!

Langfuse Cloud or Self-Hosted?

Langfuse Cloud

If Self-Hosted

No response

If Langfuse Cloud

https://cloud.langfuse.com/project/cmivtj9v8009dad07heea7du0

SDK and integration versions

$ java --version
No version is set for command java
Consider adding one of the following versions in your config file at /Users/e.laskaridis/Projects/sandbox/mcp-playground/.tool-versions
java temurin-21.0.8+9.0.LTS
java temurin-21.0.2+13.0.LTS

plugins {
	kotlin("jvm") version "1.9.25"
	kotlin("plugin.spring") version "1.9.25"
	id("org.springframework.boot") version "3.5.7"
	id("io.spring.dependency-management") version "1.1.7"
}

Pre-Submission Checklist

• I have checked for existing issues/discussions and consulted Langfuse AI.

[inkeep-ai-assistant[bot]]

The 404 error you're seeing is expected behavior. Langfuse only supports trace ingestion via OpenTelemetry—logs and metrics are not supported at the OTLP endpoint(1). The error message about failing to export logs can be safely ignored as it doesn't affect tracing functionality(1).

However, there are two configuration issues to address:

1. Use the correct endpoint for Spring AI

Your environment variable should point to the base OTLP endpoint, not the traces-specific path(2)(3):

OTEL_EXPORTER_OTLP_ENDPOINT=https://cloud.langfuse.com/api/public/otel

If your Spring Boot setup requires signal-specific variables, use(3):

OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=https://cloud.langfuse.com/api/public/otel/v1/traces

2. Fix your authentication header

Your OTEL_EXPORTER_OTLP_HEADERS contains "Authorization=Basic" twice. It should be(2)(3):

OTEL_EXPORTER_OTLP_HEADERS=Authorization=Basic <base64_encoded_credentials>

The base64 encoding should include both your public and secret keys in the format public_key:secret_key(3). Your current command only encodes the public key. Use:

echo -n "pk-lf-...:sk-lf-..." | base64

3. Add the custom span processor

The Spring AI integration requires a custom ChatModelCompletionContentObservationFilter to properly capture prompt and completion content in spans(2). Without this component, your traces may be incomplete. The full implementation is provided in the Spring AI integration guide(2).

After making these corrections, your traces should appear in Langfuse without 404 errors for the trace data itself (though log-related 404s may still appear and can be ignored).

📚 Sources:

1. User SpringAI + langfuse failed to export logs
2. Integrating Langfuse with Spring AI
3. LLM Observability via OpenTelemetry

---

Have another question? Just tag @inkeep.

[laskaridis]

1. I've used the correct endpoint as suggested

OTEL_EXPORTER_OTLP_ENDPOINT=https://cloud.langfuse.com/api/public/otel

2. Confirmed that my auth header is set according to the instructions

OTEL_EXPORTER_OTLP_HEADERS=Authorization=Basic <base64_encoded_credentials>

Generated using both my public and private keys as follows:

echo -n "pk-lf-...:sk-lf-..." | base64

3. I have added the ChatModelCompletionContentObservationFilter as suggested in the instructions

Unfortunately i'm still getting the following error and don't see any traces in my project:

[angfuse.com/...] [                                                 ] i.o.exporter.internal.http.HttpExporter  : Failed to export logs. Server responded with HTTP status code 404. Error message: Unable to parse response body, HTTP status message: 

@inkeep

[inkeep-ai-assistant[bot]]

The 404 error for logs is expected and can be ignored(1). Langfuse does not support OpenTelemetry metrics or logs—only traces are ingested at the /api/public/otel endpoint(2)(1). The error message about failing to export logs does not affect trace functionality(1).

Since you're not seeing traces in your project despite correct configuration, here are additional troubleshooting steps:

1. Verify protocol configuration

Langfuse does not support gRPC for the OpenTelemetry endpoint—you must use HTTP/protobuf(2). Add this to your application.properties:

management.otlp.tracing.endpoint=https://cloud.langfuse.com/api/public/otel/v1/traces
otel.exporter.otlp.protocol=http/protobuf

(3)

2. Confirm Spring AI observations are enabled

Ensure these properties are set in application.properties:

spring.ai.chat.observations.log-prompt=true
spring.ai.chat.observations.log-completion=true
management.tracing.sampling.probability=1.0

(3)

3. Check for mixed environment variable usage

When using Spring Boot's OpenTelemetry starter, environment variables and application properties can conflict. Try setting the endpoint and headers only in application.properties instead of environment variables:

management.otlp.tracing.endpoint=https://cloud.langfuse.com/api/public/otel/v1/traces
management.otlp.tracing.headers.Authorization=Basic <your_base64_credentials>

(3)

4. Verify the ChatModelCompletionContentObservationFilter bean

The custom observation filter must be registered as a Spring bean to capture prompt and completion content properly(3). Confirm your implementation matches the documentation exactly.

5. Test with a minimal example

Try a simple chat call to verify traces are being generated. If you still don't see traces after addressing the above points, the issue may be with how Spring Boot's OpenTelemetry auto-configuration is initializing(3).

The log export 404 errors will continue to appear but should not prevent trace data from reaching Langfuse(1).

📚 Sources:

1. bug: POST /api/public/otel/v1/metrics HTTP/1.1" 404 None
2. LLM Observability via OpenTelemetry
3. Integrating Langfuse with Spring AI

---

Have another question? Just tag @inkeep.

[laskaridis]

@inkeep

1. Added the following config in application.properties:

spring.ai.openai.api-key=...
spring.ai.chat.observations.log-prompt=true
spring.ai.chat.observations.log-completion=true

management.tracing.sampling.probability=1.0
management.observations.annotations.enabled=true
management.otlp.tracing.endpoint=https://cloud.langfuse.com/api/public/otel/v1/traces
management.otlp.tracing.headers.Authorization=Basic <...>

otel.exporter.otlp.protocol=http/protobuf

2. Removed the env variables in favour of properties as suggested

I now get the following:

java.net.ConnectException: Failed to connect to localhost/[0:0:0:0:0:0:0:1]:4318
	at okhttp3.internal.connection.RealConnection.connectSocket(RealConnection.kt:297) ~[okhttp-4.12.0.jar:na]
	at okhttp3.internal.connection.RealConnection.connect(RealConnection.kt:207) ~[okhttp-4.12.0.jar:na]
	...

I'm using the cloud deployment.

[inkeep-ai-assistant[bot]]

The connection error to localhost:4318 indicates that Spring Boot's OpenTelemetry auto-configuration is ignoring your management.otlp.tracing.endpoint property and falling back to the default local endpoint(1).

Use environment variables instead of Spring properties

Based on the Spring AI integration documentation and community troubleshooting, the recommended approach is to use OpenTelemetry environment variables rather than Spring Boot management properties(2)(1):

OTEL_EXPORTER_OTLP_ENDPOINT=https://cloud.langfuse.com/api/public/otel
OTEL_EXPORTER_OTLP_HEADERS=Authorization=Basic <your_base64_credentials>

The endpoint should be the base OTLP path (/api/public/otel), not the traces-specific path(2)(3).

Verify your dependencies match the documentation

Confirm your build.gradle.kts includes exactly these dependencies as specified in the integration guide(2):

dependencies {
    implementation("org.springframework.boot:spring-boot-starter-actuator")
    implementation("org.springframework.ai:spring-ai-starter-model-openai")
    implementation("io.opentelemetry.instrumentation:opentelemetry-spring-boot-starter")
    implementation("io.opentelemetry:opentelemetry-exporter-otlp")
    implementation("io.micrometer:micrometer-tracing-bridge-otel")
}

dependencyManagement {
    imports {
        mavenBom("org.springframework.ai:spring-ai-bom:1.0.0")
        mavenBom("io.opentelemetry.instrumentation:opentelemetry-instrumentation-bom:2.17.0")
    }
}

(2)

Keep your application.properties minimal

Your application.properties should only contain(2):

spring.ai.openai.api-key=<your_key>
spring.ai.chat.observations.log-prompt=true
spring.ai.chat.observations.log-completion=true
management.tracing.sampling.probability=1.0
management.observations.annotations.enabled=true

Remove the management.otlp.* and otel.exporter.* properties—these conflict with the environment variables(1).

Common issue from the community

A user reported this exact same error pattern and confirmed that switching to environment variables resolved the connection issue(1). The Spring Boot OpenTelemetry starter appears to prioritize environment variables over application properties for OTLP configuration.

After making these changes, restart your application and the traces should flow to Langfuse Cloud. The 404 log errors will still appear but can be ignored as Langfuse only supports trace ingestion via OpenTelemetry(4).

📚 Sources:

1. Spring AI w/ Langfuse via OpenTelemetry not quite working
2. Integrating Langfuse with Spring AI
3. LLM Observability via OpenTelemetry
4. User SpringAI + langfuse failed to export logs

---

Have another question? Just tag @inkeep.