Mnovich/otlp v2 (#3772)

2025-08-01 12:15:12 -07:00
parent 1a8b3b1a42
commit 3b6a92b715
8 changed files with 602 additions and 35 deletions
@@ -992,13 +992,40 @@ dependencies = [
 "tracing",
 ]

+[[package]]
+name = "axum"
+version = "0.7.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "edca88bc138befd0323b20752846e6587272d3b03b0343c8ea28a6f819e6e71f"
+dependencies = [
+ "async-trait",
+ "axum-core 0.4.5",
+ "bytes",
+ "futures-util",
+ "http 1.2.0",
+ "http-body 1.0.1",
+ "http-body-util",
+ "itoa",
+ "matchit 0.7.3",
+ "memchr",
+ "mime",
+ "percent-encoding",
+ "pin-project-lite",
+ "rustversion",
+ "serde",
+ "sync_wrapper 1.0.2",
+ "tower 0.5.2",
+ "tower-layer",
+ "tower-service",
+]
+
 [[package]]
 name = "axum"
 version = "0.8.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6d6fd624c75e18b3b4c6b9caf42b1afe24437daaee904069137d8bab077be8b8"
 dependencies = [
- "axum-core",
+ "axum-core 0.5.0",
 "axum-macros",
 "base64 0.22.1",
 "bytes",
@@ -1010,7 +1037,7 @@ dependencies = [
 "hyper 1.6.0",
 "hyper-util",
 "itoa",
- "matchit",
+ "matchit 0.8.4",
 "memchr",
 "mime",
 "percent-encoding",
@@ -1030,6 +1057,26 @@ dependencies = [
 "tracing",
 ]

+[[package]]
+name = "axum-core"
+version = "0.4.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "09f2bd6146b97ae3359fa0cc6d6b376d9539582c7b4220f041a33ec24c226199"
+dependencies = [
+ "async-trait",
+ "bytes",
+ "futures-util",
+ "http 1.2.0",
+ "http-body 1.0.1",
+ "http-body-util",
+ "mime",
+ "pin-project-lite",
+ "rustversion",
+ "sync_wrapper 1.0.2",
+ "tower-layer",
+ "tower-service",
+]
+
 [[package]]
 name = "axum-core"
 version = "0.5.0"
@@ -2695,7 +2742,7 @@ dependencies = [
 "futures",
 "hyper 0.14.32",
 "hyper-rustls 0.24.2",
- "hyper-timeout",
+ "hyper-timeout 0.4.1",
 "log",
 "pin-project",
 "rand 0.8.5",
@@ -3246,7 +3293,7 @@ dependencies = [
 "aws-sdk-bedrockruntime",
 "aws-sdk-sagemakerruntime",
 "aws-smithy-types",
- "axum",
+ "axum 0.8.1",
 "base64 0.21.7",
 "blake3",
 "chrono",
@@ -3271,6 +3318,9 @@ dependencies = [
 "mockall",
 "nanoid",
 "once_cell",
+ "opentelemetry",
+ "opentelemetry-otlp",
+ "opentelemetry_sdk",
 "rand 0.8.5",
 "regex",
 "reqwest 0.12.12",
@@ -3289,7 +3339,9 @@ dependencies = [
 "tokio-cron-scheduler",
 "tokio-stream",
 "tokio-util",
+ "tonic",
 "tracing",
+ "tracing-opentelemetry",
 "tracing-subscriber",
 "url",
 "urlencoding",
@@ -3329,7 +3381,7 @@ version = "1.1.0"
 dependencies = [
 "anyhow",
 "async-trait",
- "axum",
+ "axum 0.8.1",
 "base64 0.22.1",
 "bat",
 "chrono",
@@ -3429,7 +3481,7 @@ version = "1.1.0"
 dependencies = [
 "anyhow",
 "async-trait",
- "axum",
+ "axum 0.8.1",
 "base64 0.21.7",
 "bytes",
 "chrono",
@@ -3799,6 +3851,19 @@ dependencies = [
 "tokio-io-timeout",
 ]

+[[package]]
+name = "hyper-timeout"
+version = "0.5.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2b90d566bffbce6a75bd8b09a05aa8c2cb1fabb6cb348f8840c9e4c90a0d83b0"
+dependencies = [
+ "hyper 1.6.0",
+ "hyper-util",
+ "pin-project-lite",
+ "tokio",
+ "tower-service",
+]
+
 [[package]]
 name = "hyper-util"
 version = "0.1.10"
@@ -4391,7 +4456,7 @@ dependencies = [
 "object_store",
 "permutation",
 "pin-project",
- "prost",
+ "prost 0.12.6",
 "prost-build",
 "rand 0.8.5",
 "roaring",
@@ -4450,7 +4515,7 @@ dependencies = [
 "num_cpus",
 "object_store",
 "pin-project",
- "prost",
+ "prost 0.12.6",
 "rand 0.8.5",
 "roaring",
 "serde_json",
@@ -4484,7 +4549,7 @@ dependencies = [
 "lance-core",
 "lazy_static",
 "log",
- "prost",
+ "prost 0.12.6",
 "snafu",
 "tokio",
 ]
@@ -4518,7 +4583,7 @@ dependencies = [
 "log",
 "num-traits",
 "paste",
- "prost",
+ "prost 0.12.6",
 "prost-build",
 "prost-types",
 "rand 0.8.5",
@@ -4555,7 +4620,7 @@ dependencies = [
 "log",
 "num-traits",
 "object_store",
- "prost",
+ "prost 0.12.6",
 "prost-build",
 "prost-types",
 "roaring",
@@ -4603,7 +4668,7 @@ dependencies = [
 "moka",
 "num-traits",
 "object_store",
- "prost",
+ "prost 0.12.6",
 "prost-build",
 "rand 0.8.5",
 "rayon",
@@ -4649,7 +4714,7 @@ dependencies = [
 "object_store",
 "path_abs",
 "pin-project",
- "prost",
+ "prost 0.12.6",
 "prost-build",
 "rand 0.8.5",
 "shellexpand",
@@ -4710,7 +4775,7 @@ dependencies = [
 "lazy_static",
 "log",
 "object_store",
- "prost",
+ "prost 0.12.6",
 "prost-build",
 "prost-types",
 "rand 0.8.5",
@@ -5065,6 +5130,12 @@ dependencies = [
 "regex-automata 0.1.10",
 ]

+[[package]]
+name = "matchit"
+version = "0.7.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0e7465ac9959cc2b1404e8e2367b43684a6d13790fe23056cc8c6c5a6b7bcb94"
+
 [[package]]
 name = "matchit"
 version = "0.8.4"
@@ -5087,7 +5158,7 @@ version = "0.1.0"
 dependencies = [
 "anyhow",
 "async-trait",
- "axum",
+ "axum 0.8.1",
 "base64 0.22.1",
 "eventsource-client",
 "futures",
@@ -5754,6 +5825,87 @@ dependencies = [
 "vcpkg",
 ]

+[[package]]
+name = "opentelemetry"
+version = "0.27.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ab70038c28ed37b97d8ed414b6429d343a8bbf44c9f79ec854f3a643029ba6d7"
+dependencies = [
+ "futures-core",
+ "futures-sink",
+ "js-sys",
+ "pin-project-lite",
+ "thiserror 1.0.69",
+ "tracing",
+]
+
+[[package]]
+name = "opentelemetry-http"
+version = "0.27.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "10a8a7f5f6ba7c1b286c2fbca0454eaba116f63bbe69ed250b642d36fbb04d80"
+dependencies = [
+ "async-trait",
+ "bytes",
+ "http 1.2.0",
+ "opentelemetry",
+ "reqwest 0.12.12",
+]
+
+[[package]]
+name = "opentelemetry-otlp"
+version = "0.27.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "91cf61a1868dacc576bf2b2a1c3e9ab150af7272909e80085c3173384fe11f76"
+dependencies = [
+ "async-trait",
+ "futures-core",
+ "http 1.2.0",
+ "opentelemetry",
+ "opentelemetry-http",
+ "opentelemetry-proto",
+ "opentelemetry_sdk",
+ "prost 0.13.5",
+ "reqwest 0.12.12",
+ "thiserror 1.0.69",
+ "tokio",
+ "tonic",
+ "tracing",
+]
+
+[[package]]
+name = "opentelemetry-proto"
+version = "0.27.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a6e05acbfada5ec79023c85368af14abd0b307c015e9064d249b2a950ef459a6"
+dependencies = [
+ "opentelemetry",
+ "opentelemetry_sdk",
+ "prost 0.13.5",
+ "tonic",
+]
+
+[[package]]
+name = "opentelemetry_sdk"
+version = "0.27.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "231e9d6ceef9b0b2546ddf52335785ce41252bc7474ee8ba05bfad277be13ab8"
+dependencies = [
+ "async-trait",
+ "futures-channel",
+ "futures-executor",
+ "futures-util",
+ "glob",
+ "opentelemetry",
+ "percent-encoding",
+ "rand 0.8.5",
+ "serde_json",
+ "thiserror 1.0.69",
+ "tokio",
+ "tokio-stream",
+ "tracing",
+]
+
 [[package]]
 name = "option-ext"
 version = "0.2.0"
@@ -6192,7 +6344,17 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "deb1435c188b76130da55f17a466d252ff7b1418b2ad3e037d127b94e3411f29"
 dependencies = [
 "bytes",
- "prost-derive",
+ "prost-derive 0.12.6",
+]
+
+[[package]]
+name = "prost"
+version = "0.13.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2796faa41db3ec313a31f7624d9286acf277b52de526150b7e69f3debf891ee5"
+dependencies = [
+ "bytes",
+ "prost-derive 0.13.5",
 ]

 [[package]]
@@ -6209,7 +6371,7 @@ dependencies = [
 "once_cell",
 "petgraph",
 "prettyplease",
- "prost",
+ "prost 0.12.6",
 "prost-types",
 "regex",
 "syn 2.0.99",
@@ -6229,13 +6391,26 @@ dependencies = [
 "syn 2.0.99",
 ]

+[[package]]
+name = "prost-derive"
+version = "0.13.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8a56d757972c98b346a9b766e3f02746cde6dd1cd1d1d563472929fdd74bec4d"
+dependencies = [
+ "anyhow",
+ "itertools 0.13.0",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.99",
+]
+
 [[package]]
 name = "prost-types"
 version = "0.12.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9091c90b0a32608e984ff2fa4091273cbdd755d54935c51d520887f4a1dbd5b0"
 dependencies = [
- "prost",
+ "prost 0.12.6",
 ]

 [[package]]
@@ -8335,6 +8510,36 @@ dependencies = [
 "winnow",
 ]

+[[package]]
+name = "tonic"
+version = "0.12.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "877c5b330756d856ffcc4553ab34a5684481ade925ecc54bcd1bf02b1d0d4d52"
+dependencies = [
+ "async-stream",
+ "async-trait",
+ "axum 0.7.9",
+ "base64 0.22.1",
+ "bytes",
+ "h2 0.4.8",
+ "http 1.2.0",
+ "http-body 1.0.1",
+ "http-body-util",
+ "hyper 1.6.0",
+ "hyper-timeout 0.5.2",
+ "hyper-util",
+ "percent-encoding",
+ "pin-project",
+ "prost 0.13.5",
+ "socket2 0.5.8",
+ "tokio",
+ "tokio-stream",
+ "tower 0.4.13",
+ "tower-layer",
+ "tower-service",
+ "tracing",
+]
+
 [[package]]
 name = "tower"
 version = "0.4.13"
@@ -8343,9 +8548,13 @@ checksum = "b8fa9be0de6cf49e536ce1851f987bd21a43b771b09473c3549a6c853db37c1c"
 dependencies = [
 "futures-core",
 "futures-util",
+ "indexmap 1.9.3",
 "pin-project",
 "pin-project-lite",
+ "rand 0.8.5",
+ "slab",
 "tokio",
+ "tokio-util",
 "tower-layer",
 "tower-service",
 "tracing",
@@ -8460,6 +8669,24 @@ dependencies = [
 "tracing-core",
 ]

+[[package]]
+name = "tracing-opentelemetry"
+version = "0.28.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "97a971f6058498b5c0f1affa23e7ea202057a7301dbff68e968b2d578bcbd053"
+dependencies = [
+ "js-sys",
+ "once_cell",
+ "opentelemetry",
+ "opentelemetry_sdk",
+ "smallvec",
+ "tracing",
+ "tracing-core",
+ "tracing-log",
+ "tracing-subscriber",
+ "web-time",
+]
+
 [[package]]
 name = "tracing-serde"
 version = "0.2.0"
@@ -701,6 +701,28 @@ pub async fn cli() -> Result<()> {
        eprintln!("Warning: Failed to update project tracker: {}", e);
    }

+    let command_name = match &cli.command {
+        Some(Command::Configure {}) => "configure",
+        Some(Command::Info { .. }) => "info",
+        Some(Command::Mcp { .. }) => "mcp",
+        Some(Command::Session { .. }) => "session",
+        Some(Command::Project {}) => "project",
+        Some(Command::Projects) => "projects",
+        Some(Command::Run { .. }) => "run",
+        Some(Command::Schedule { .. }) => "schedule",
+        Some(Command::Update { .. }) => "update",
+        Some(Command::Bench { .. }) => "bench",
+        Some(Command::Recipe { .. }) => "recipe",
+        Some(Command::Web { .. }) => "web",
+        None => "default_session",
+    };
+
+    tracing::info!(
+        monotonic_counter.goose.cli_commands = 1,
+        command = command_name,
+        "CLI command executed"
+    );
+
    match cli.command {
        Some(Command::Configure {}) => {
            let _ = handle_configure().await;
@@ -875,6 +897,11 @@ pub async fn cli() -> Result<()> {
                    (input_config, None)
                }
                (_, _, Some(recipe_name)) => {
+                    tracing::info!(monotonic_counter.goose.recipe_runs = 1,
+                        recipe_name = %recipe_name,
+                        "Recipe execution started"
+                    );
+
                    if explain {
                        explain_recipe(&recipe_name, params)?;
                        return Ok(());
@@ -11,7 +11,7 @@ use tracing_subscriber::{
    Registry,
 };

-use goose::tracing::langfuse_layer;
+use goose::tracing::{langfuse_layer, otlp_layer};
 use goose_bench::bench_session::BenchAgentError;
 use goose_bench::error_capture::ErrorCaptureLayer;

@@ -140,7 +140,21 @@ fn setup_logging_internal(
                layers.push(ErrorCaptureLayer::new().boxed());
            }

-            // Add Langfuse layer if available
+            if !force {
+                if let Ok((otlp_tracing_layer, otlp_metrics_layer)) = otlp_layer::init_otlp() {
+                    layers.push(
+                        otlp_tracing_layer
+                            .with_filter(otlp_layer::create_otlp_tracing_filter())
+                            .boxed(),
+                    );
+                    layers.push(
+                        otlp_metrics_layer
+                            .with_filter(otlp_layer::create_otlp_metrics_filter())
+                            .boxed(),
+                    );
+                }
+            }
+
            if let Some(langfuse) = langfuse_layer::create_langfuse_observer() {
                layers.push(langfuse.with_filter(LevelFilter::DEBUG).boxed());
            }
@@ -1009,6 +1009,17 @@ impl Session {
                            }
                            // otherwise we have a model/tool to render
                            else {
+                                for content in &message.content {
+                                    if let MessageContent::ToolRequest(tool_request) = content {
+                                        if let Ok(tool_call) = &tool_request.tool_call {
+                                            tracing::info!(monotonic_counter.goose.tool_calls = 1,
+                                                tool_name = %tool_call.name,
+                                                "Tool call executed"
+                                            );
+                                        }
+                                    }
+                                }
+
                                push_message(&mut self.messages, message.clone());

                                // No need to update description on assistant messages
@@ -9,7 +9,7 @@ use tracing_subscriber::{
 };

 use goose::config::APP_STRATEGY;
-use goose::tracing::langfuse_layer;
+use goose::tracing::{langfuse_layer, otlp_layer};

 /// Returns the directory where log files should be stored.
 /// Creates the directory structure if it doesn't exist.
@@ -91,22 +91,33 @@ pub fn setup_logging(name: Option<&str>) -> Result<()> {
            .add_directive(LevelFilter::WARN.into())
    });

-    // Build the subscriber with required layers
-    let subscriber = Registry::default()
-        .with(file_layer.with_filter(env_filter))
-        .with(console_layer.with_filter(LevelFilter::INFO));
+    let mut layers = vec![
+        file_layer.with_filter(env_filter).boxed(),
+        console_layer.with_filter(LevelFilter::INFO).boxed(),
+    ];

-    // Initialize with Langfuse if available
-    if let Some(langfuse) = langfuse_layer::create_langfuse_observer() {
-        subscriber
-            .with(langfuse.with_filter(LevelFilter::DEBUG))
-            .try_init()
-            .context("Failed to set global subscriber")?;
-    } else {
-        subscriber
-            .try_init()
-            .context("Failed to set global subscriber")?;
+    if let Ok((otlp_tracing_layer, otlp_metrics_layer)) = otlp_layer::init_otlp() {
+        layers.push(
+            otlp_tracing_layer
+                .with_filter(otlp_layer::create_otlp_tracing_filter())
+                .boxed(),
+        );
+        layers.push(
+            otlp_metrics_layer
+                .with_filter(otlp_layer::create_otlp_metrics_filter())
+                .boxed(),
+        );
    }

+    if let Some(langfuse) = langfuse_layer::create_langfuse_observer() {
+        layers.push(langfuse.with_filter(LevelFilter::DEBUG).boxed());
+    }
+
+    let subscriber = Registry::default().with(layers);
+
+    subscriber
+        .try_init()
+        .context("Failed to set global subscriber")?;
+
    Ok(())
 }
@@ -58,6 +58,11 @@ webbrowser = "0.8"
 lazy_static = "1.5.0"
 tracing = "0.1"
 tracing-subscriber = "0.3"
+tracing-opentelemetry = "0.28"
+opentelemetry = "0.27"
+opentelemetry_sdk = { version = "0.27", features = ["rt-tokio", "metrics"] }
+opentelemetry-otlp = { version = "0.27", features = ["grpc-tonic", "http-proto", "reqwest-client"] }
+tonic = "0.12"
 keyring = { version = "3.6.2", features = ["apple-native", "windows-native", "sync-secret-service", "vendored"] }
 serde_yaml = "0.9.34"
 once_cell = "1.20.2"
@@ -1,7 +1,12 @@
 pub mod langfuse_layer;
 mod observation_layer;
+pub mod otlp_layer;

 pub use langfuse_layer::{create_langfuse_observer, LangfuseBatchManager};
 pub use observation_layer::{
    flatten_metadata, map_level, BatchManager, ObservationLayer, SpanData, SpanTracker,
 };
+pub use otlp_layer::{
+    create_otlp_metrics_filter, create_otlp_tracing_filter, create_otlp_tracing_layer,
+    init_otlp_metrics, init_otlp_tracing, init_otlp_tracing_only, shutdown_otlp, OtlpConfig,
+};
@@ -0,0 +1,267 @@
+use opentelemetry::trace::TracerProvider;
+use opentelemetry::{global, KeyValue};
+use opentelemetry_otlp::WithExportConfig;
+use opentelemetry_sdk::trace::{self, RandomIdGenerator, Sampler};
+use opentelemetry_sdk::{runtime, Resource};
+use std::env;
+use std::time::Duration;
+use tracing::{Level, Metadata};
+use tracing_opentelemetry::{MetricsLayer, OpenTelemetryLayer};
+use tracing_subscriber::filter::FilterFn;
+
+pub type OtlpTracingLayer =
+    OpenTelemetryLayer<tracing_subscriber::Registry, opentelemetry_sdk::trace::Tracer>;
+pub type OtlpMetricsLayer = MetricsLayer<tracing_subscriber::Registry>;
+pub type OtlpLayers = (OtlpTracingLayer, OtlpMetricsLayer);
+pub type OtlpResult<T> = Result<T, Box<dyn std::error::Error + Send + Sync>>;
+
+#[derive(Debug, Clone)]
+pub struct OtlpConfig {
+    pub endpoint: String,
+    pub timeout: Duration,
+}
+
+impl Default for OtlpConfig {
+    fn default() -> Self {
+        Self {
+            endpoint: "http://localhost:4318".to_string(),
+            timeout: Duration::from_secs(10),
+        }
+    }
+}
+
+impl OtlpConfig {
+    pub fn from_env() -> Option<Self> {
+        if let Ok(endpoint) = env::var("OTEL_EXPORTER_OTLP_ENDPOINT") {
+            let mut config = Self {
+                endpoint,
+                timeout: Duration::from_secs(10),
+            };
+
+            if let Ok(timeout_str) = env::var("OTEL_EXPORTER_OTLP_TIMEOUT") {
+                if let Ok(timeout_ms) = timeout_str.parse::<u64>() {
+                    config.timeout = Duration::from_millis(timeout_ms);
+                }
+            }
+
+            Some(config)
+        } else {
+            None
+        }
+    }
+}
+
+pub fn init_otlp_tracing(config: &OtlpConfig) -> OtlpResult<()> {
+    let resource = Resource::new(vec![
+        KeyValue::new("service.name", "goose"),
+        KeyValue::new("service.version", env!("CARGO_PKG_VERSION")),
+        KeyValue::new("service.namespace", "goose"),
+    ]);
+
+    let exporter = opentelemetry_otlp::SpanExporter::builder()
+        .with_http()
+        .with_endpoint(&config.endpoint)
+        .with_timeout(config.timeout)
+        .build()?;
+
+    let tracer_provider = trace::TracerProvider::builder()
+        .with_batch_exporter(exporter, runtime::Tokio)
+        .with_resource(resource.clone())
+        .with_id_generator(RandomIdGenerator::default())
+        .with_sampler(Sampler::AlwaysOn)
+        .build();
+
+    global::set_tracer_provider(tracer_provider);
+
+    Ok(())
+}
+
+pub fn init_otlp_metrics(config: &OtlpConfig) -> OtlpResult<()> {
+    let resource = Resource::new(vec![
+        KeyValue::new("service.name", "goose"),
+        KeyValue::new("service.version", env!("CARGO_PKG_VERSION")),
+        KeyValue::new("service.namespace", "goose"),
+    ]);
+
+    let exporter = opentelemetry_otlp::MetricExporter::builder()
+        .with_http()
+        .with_endpoint(&config.endpoint)
+        .with_timeout(config.timeout)
+        .build()?;
+
+    let meter_provider = opentelemetry_sdk::metrics::SdkMeterProvider::builder()
+        .with_resource(resource)
+        .with_reader(
+            opentelemetry_sdk::metrics::PeriodicReader::builder(exporter, runtime::Tokio)
+                .with_interval(Duration::from_secs(3))
+                .build(),
+        )
+        .build();
+
+    global::set_meter_provider(meter_provider);
+
+    Ok(())
+}
+
+pub fn create_otlp_tracing_layer() -> OtlpResult<OtlpTracingLayer> {
+    let config =
+        OtlpConfig::from_env().ok_or("OTEL_EXPORTER_OTLP_ENDPOINT environment variable not set")?;
+
+    let resource = Resource::new(vec![
+        KeyValue::new("service.name", "goose"),
+        KeyValue::new("service.version", env!("CARGO_PKG_VERSION")),
+        KeyValue::new("service.namespace", "goose"),
+    ]);
+
+    let exporter = opentelemetry_otlp::SpanExporter::builder()
+        .with_http()
+        .with_endpoint(&config.endpoint)
+        .with_timeout(config.timeout)
+        .build()?;
+
+    let tracer_provider = trace::TracerProvider::builder()
+        .with_batch_exporter(exporter, runtime::Tokio)
+        .with_resource(resource)
+        .with_id_generator(RandomIdGenerator::default())
+        .with_sampler(Sampler::AlwaysOn)
+        .build();
+
+    let tracer = tracer_provider.tracer("goose");
+    Ok(tracing_opentelemetry::layer().with_tracer(tracer))
+}
+
+pub fn create_otlp_metrics_layer() -> OtlpResult<OtlpMetricsLayer> {
+    let config =
+        OtlpConfig::from_env().ok_or("OTEL_EXPORTER_OTLP_ENDPOINT environment variable not set")?;
+
+    let resource = Resource::new(vec![
+        KeyValue::new("service.name", "goose"),
+        KeyValue::new("service.version", env!("CARGO_PKG_VERSION")),
+        KeyValue::new("service.namespace", "goose"),
+    ]);
+
+    let exporter = opentelemetry_otlp::MetricExporter::builder()
+        .with_http()
+        .with_endpoint(&config.endpoint)
+        .with_timeout(config.timeout)
+        .build()?;
+
+    let meter_provider = opentelemetry_sdk::metrics::SdkMeterProvider::builder()
+        .with_resource(resource)
+        .with_reader(
+            opentelemetry_sdk::metrics::PeriodicReader::builder(exporter, runtime::Tokio)
+                .with_interval(Duration::from_secs(5)) // Reduced from 30s to 5s for faster metrics
+                .build(),
+        )
+        .build();
+
+    global::set_meter_provider(meter_provider.clone());
+
+    Ok(tracing_opentelemetry::MetricsLayer::new(meter_provider))
+}
+
+pub fn init_otlp() -> OtlpResult<OtlpLayers> {
+    let tracing_layer = create_otlp_tracing_layer()?;
+    let metrics_layer = create_otlp_metrics_layer()?;
+    Ok((tracing_layer, metrics_layer))
+}
+
+pub fn init_otlp_tracing_only() -> OtlpResult<OtlpTracingLayer> {
+    create_otlp_tracing_layer()
+}
+
+/// Creates a custom filter for OTLP tracing that captures:
+/// - All spans at INFO level and above
+/// - Specific spans marked with "otel.trace" field
+/// - Events from specific modules related to telemetry
+pub fn create_otlp_tracing_filter() -> FilterFn<impl Fn(&Metadata<'_>) -> bool> {
+    FilterFn::new(|metadata: &Metadata<'_>| {
+        if metadata.level() <= &Level::INFO {
+            return true;
+        }
+
+        if metadata.level() == &Level::DEBUG {
+            let target = metadata.target();
+            if target.starts_with("goose::")
+                || target.starts_with("opentelemetry")
+                || target.starts_with("tracing_opentelemetry")
+            {
+                return true;
+            }
+        }
+
+        false
+    })
+}
+
+/// Creates a custom filter for OTLP metrics that captures:
+/// - All events at INFO level and above
+/// - Specific events marked with "otel.metric" field
+/// - Events that should be converted to metrics
+pub fn create_otlp_metrics_filter() -> FilterFn<impl Fn(&Metadata<'_>) -> bool> {
+    FilterFn::new(|metadata: &Metadata<'_>| {
+        if metadata.level() <= &Level::INFO {
+            return true;
+        }
+
+        if metadata.level() == &Level::DEBUG {
+            let target = metadata.target();
+            if target.starts_with("goose::telemetry")
+                || target.starts_with("goose::metrics")
+                || target.contains("metric")
+            {
+                return true;
+            }
+        }
+
+        false
+    })
+}
+
+/// Shutdown OTLP providers gracefully
+pub fn shutdown_otlp() {
+    global::shutdown_tracer_provider();
+
+    // Note: There's currently no clean way to shutdown the global meter provider
+    // in the OpenTelemetry Rust SDK. The meter provider will be cleaned up when
+    // the process exits. Individual meter providers can be shut down if you have
+    // a direct reference to them.
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::env;
+
+    #[test]
+    fn test_otlp_config_default() {
+        let config = OtlpConfig::default();
+        assert_eq!(config.endpoint, "http://localhost:4318");
+        assert_eq!(config.timeout, Duration::from_secs(10));
+    }
+
+    #[test]
+    fn test_otlp_config_from_env() {
+        let original_endpoint = env::var("OTEL_EXPORTER_OTLP_ENDPOINT").ok();
+        let original_timeout = env::var("OTEL_EXPORTER_OTLP_TIMEOUT").ok();
+
+        env::remove_var("OTEL_EXPORTER_OTLP_ENDPOINT");
+        assert!(OtlpConfig::from_env().is_none());
+
+        env::set_var("OTEL_EXPORTER_OTLP_ENDPOINT", "http://test:4317");
+        env::set_var("OTEL_EXPORTER_OTLP_TIMEOUT", "5000");
+
+        let config = OtlpConfig::from_env().unwrap();
+        assert_eq!(config.endpoint, "http://test:4317");
+        assert_eq!(config.timeout, Duration::from_millis(5000));
+
+        match original_endpoint {
+            Some(val) => env::set_var("OTEL_EXPORTER_OTLP_ENDPOINT", val),
+            None => env::remove_var("OTEL_EXPORTER_OTLP_ENDPOINT"),
+        }
+        match original_timeout {
+            Some(val) => env::set_var("OTEL_EXPORTER_OTLP_TIMEOUT", val),
+            None => env::remove_var("OTEL_EXPORTER_OTLP_TIMEOUT"),
+        }
+    }
+}