Add example for logging to multiple projects with OTEL tracing

Appraisals

@@ -49,3 +49,9 @@ appraise "opentelemetry-latest" do
   gem "opentelemetry-sdk", ">= 1.10"
   gem "opentelemetry-exporter-otlp", ">= 0.31"
 end
+
+# for multiple_projects.rb only, test both openai and ruby_llm
+appraise "ruby-llm-openai" do
+  gem "openai", ">= 0.34"
+  gem "ruby_llm", ">= 1.9"
+end

Gemfile.lock

@@ -2,6 +2,7 @@ PATH
   remote: .
   specs:
     braintrust (0.0.11)
+      base64 (~> 0.2)
       openssl (~> 3.3.1)
       opentelemetry-exporter-otlp (~> 0.28)
       opentelemetry-sdk (~> 1.3)
@@ -50,7 +51,7 @@ GEM
       builder
       minitest (>= 5.0)
       ruby-progressbar
-    openssl (3.3.1)
+    openssl (3.3.2)
     opentelemetry-api (1.7.0)
     opentelemetry-common (0.23.0)
       opentelemetry-api (~> 1.0)
@@ -153,4 +154,4 @@ DEPENDENCIES
   yard (~> 0.9)
 
 BUNDLED WITH
-   2.4.19
+  2.4.19

Rakefile

@@ -32,6 +32,7 @@ end
 
 def appraisal_for(example)
   case example
+  when /multiple_projects/ then "ruby-llm-openai"
   when /ruby_llm/ then "ruby_llm"
   when /ruby-openai/, /ruby_openai/, /alexrudall/ then "ruby-openai"
   when /anthropic/ then "anthropic"

braintrust.gemspec

@@ -31,6 +31,10 @@ Gem::Specification.new do |spec|
   # Runtime dependencies
   spec.add_runtime_dependency "opentelemetry-sdk", "~> 1.3"
   spec.add_runtime_dependency "opentelemetry-exporter-otlp", "~> 0.28"
+  # Ruby 3.4+ considers this a bundled gem, removed from default gems
+  # bundler should use default base64 lib in Ruby <3.4
+  # https://stdgems.org/base64/
+  spec.add_runtime_dependency "base64", "~> 0.2"
 
   # OpenSSL 3.3.1+ fixes macOS CRL (Certificate Revocation List) verification issues
   # that occur with OpenSSL 3.6 + Ruby (certificate verify failed: unable to get certificate CRL).

examples/README.md

@@ -14,24 +14,57 @@ export BRAINTRUST_API_KEY="your-api-key-here"
 
 ## Running Examples
 
+### Using Rake (Recommended)
+
+The rake task automatically uses the correct gemfile for each example:
+
+```bash
+# Run a single example
+rake 'example[examples/trace/multiple_projects.rb]'
+
+# Run all examples
+rake examples
+```
+
+### Running Directly
+
 From the project root:
 
 ```bash
 # Run a specific example
-ruby examples/login/login_basic.rb
+ruby examples/login.rb
 
 # Enable debug logging
-BRAINTRUST_DEBUG=true ruby examples/login/login_basic.rb
+BRAINTRUST_DEBUG=true ruby examples/login.rb
 ```
 
 ## Available Examples
 
 ### Login Examples
 
-- **`login/login_basic.rb`**: Basic login example showing how to authenticate and retrieve organization information
+- **`login.rb`**: Basic login example showing how to authenticate and retrieve organization information
+
+
+### Tracing Examples
+
+- **`trace.rb`**: Basic OpenTelemetry tracing example
+- **`trace/span_filtering.rb`**: Example of filtering out non-AI spans in traces to reduce noise
+- **`trace/trace_attachments.rb`**: Example of adding attachments (images, PDFs, BLOBs) to traces
+- **`trace/multiple_projects.rb`**: Example of logging traces to multiple Braintrust projects simultaneously
+
+### LLM Integration Examples
+
+- **`openai.rb`**: OpenAI integration example
+- **`anthropic.rb`**: Anthropic integration example
+- **`ruby_llm.rb`**: Ruby LLM integration example
+- **`alexrudall_openai.rb`**: Alexrudall's ruby-openai gem integration example
+
+### Evaluation Examples
+
+- **`eval.rb`**: Defining scorers and running evals
+- **`eval/dataset.rb`**: Running an evaluation against a dataset
+- **`eval/remote_functions.rb`**: Using remote functions (server-side prompts) in evaluations
 
-## Coming Soon
+### API Examples
 
-- OpenTelemetry tracing examples
-- OpenAI integration examples
-- Eval framework examples
+- **`api/dataset.rb`**: Dataset API usage example

examples/trace/multiple_projects.rb

@@ -0,0 +1,200 @@
+#!/usr/bin/env ruby
+# frozen_string_literal: true
+
+require "bundler/setup"
+require "braintrust"
+require "opentelemetry/sdk"
+require "ruby_llm"
+require "openai"
+
+project1 = "Project-A"
+project2 = "Project-B"
+model1 = "gpt-4o-mini"
+model2 = "claude-sonnet-4"
+
+# check for API keys
+unless ENV["OPENAI_API_KEY"] && ENV["ANTHROPIC_API_KEY"]
+  puts "Error: Both OPENAI_API_KEY and ANTHROPIC_API_KEY environment variables are required"
+  puts "Get your API key from: https://platform.openai.com/api-keys"
+  puts "Get your Anthropic API key from: https://console.anthropic.com/"
+  puts "Set with `export OPENAI_API_KEY=<your_key> and export ANTHROPIC_API_KEY=<your_key>`"
+  exit 1
+end
+
+unless ENV["BRAINTRUST_API_KEY"]
+  puts "Error: BRAINTRUST_API_KEY environment variable is required"
+  puts "Get your API key from https://www.braintrust.dev/app/settings or ask your org administrator"
+  exit 1
+end
+
+# Example: Log/Trace to Multiple Projects with Separate States
+#
+# This example demonstrates how to:
+# 1. Create multiple Braintrust states for different projects
+# 2. Set up separate tracer providers for each project
+# 3. Log traces to different projects simultaneously
+#
+# Usage:
+#   bundle exec ruby examples/trace/multiple_projects.rb
+
+# Create first state for Project A (non-global)
+state_a = Braintrust.init(
+  default_project: project1,
+  set_global: false,
+  enable_tracing: false,  # We'll manually set up tracing
+  blocking_login: true    # Ensure login completes before tracing setup
+  # Not required if only tracing, login is async by default and can lead to a broken permalink if not synchronous
+)
+# Create second state for Project B (non-global)
+state_b = Braintrust.init(
+  default_project: project2,
+  set_global: false,
+  enable_tracing: false,
+  blocking_login: true
+)
+
+# Wrap all instances of RubyLLM client
+Braintrust::Trace::Contrib::Github::Crmne::RubyLLM.wrap
+
+RubyLLM.configure do |config|
+  config.openai_api_key = ENV["OPENAI_API_KEY"]
+  config.anthropic_api_key = ENV["ANTHROPIC_API_KEY"]
+end
+
+chat_openai = RubyLLM.chat(model: model1)
+chat_anthropic = RubyLLM.chat(model: model2)
+
+# Create first tracer provider
+tracer_provider_a = OpenTelemetry::SDK::Trace::TracerProvider.new
+
+# Setup using Trace.setup
+# When you pass an explicit tracer_provider, it won't set it as global
+Braintrust::Trace.setup(state_a, tracer_provider_a)
+
+# Get tracer for Project A
+tracer_a = tracer_provider_a.tracer("MultiTurn")
+
+# Note: You can also use Trace.enable instead of Trace.setup:
+# Braintrust::Trace.enable(tracer_provider_a, state: state_a)
+# Braintrust::Trace.enable(tracer_provider_b, state: state_b)
+# Both work the same when you provide explicit providers
+
+# Now create spans in first project
+puts "\nProject A: Multi-turn conversation"
+puts "=" * 50
+root_span_a = nil
+tracer_a.in_span("chat_ask") do |span|
+  root_span_a = span
+  span.set_attribute("project", project1)
+
+  # Nested spans for multi-turn convo
+  tracer_a.in_span("turn1") do |nested_t1|
+    # Using OTEL GenAI Semantic Conventions for properties
+    # https://www.braintrust.dev/docs/integrations/sdk-integrations/opentelemetry#manual-tracing
+    # Braintrust automatically maps `gen_ai.*` attributes to native Braintrust fields
+    # tracer_b will use native fields
+    nested_t1.set_attribute("gen_ai.operation.name", "chat")
+    nested_t1.set_attribute("gen_ai.request.model", model1)
+    input = "What is the best season to visit Japan?"
+    puts "\nTurn 1 (#{model1}):"
+    puts "Q: #{input}"
+    output = chat_openai.ask(input)
+
+    nested_t1.set_attribute("gen_ai.prompt", input)
+    nested_t1.set_attribute("gen_ai.completion", output.content)
+    puts "A: #{output.content[0..100]}..."
+    puts "  Tokens: #{output.to_h[:input_tokens]} in, #{output.to_h[:output_tokens]} out"
+
+    tracer_a.in_span("turn2") do |nested_t2|
+      nested_t2.set_attribute("gen_ai.operation.name", "chat")
+      nested_t2.set_attribute("gen_ai.request.model", model2)
+      input = "Which airlines fly to Japan from SFO?"
+      puts "\nTurn 2 (#{model2}):"
+      puts "Q: #{input}"
+      output = chat_anthropic.ask(input)
+
+      nested_t2.set_attribute("gen_ai.prompt", input)
+      nested_t2.set_attribute("gen_ai.completion", output.content)
+      puts "A: #{output.content[0..100]}..."
+      puts "  Tokens: #{output.to_h[:input_tokens]} in, #{output.to_h[:output_tokens]} out"
+    end
+  end
+end
+
+puts "\n✓ Multi-turn conversation completed"
+puts "\n✓ View Project A trace in Braintrust:"
+puts "  #{Braintrust::Trace.permalink(root_span_a)}"
+
+url = "https://upload.wikimedia.org/wikipedia/commons/thumb/6/65/Tokyo_Tower_during_daytime.jpg/330px-Tokyo_Tower_during_daytime.jpg"
+
+# For second project, we'll use the Ruby OpenAI client
+# You can log to multiple projects even if your clients use different client libs
+client = OpenAI::Client.new(api_key: ENV["OPENAI_API_KEY"])
+
+# Create second tracer provider
+tracer_provider_b = OpenTelemetry::SDK::Trace::TracerProvider.new
+Braintrust::Trace.setup(state_b, tracer_provider_b)
+
+# Get tracer for Project A
+tracer_b = tracer_provider_b.tracer("ImageUpload")
+
+# Wrapping OpenAI client with second trace provider
+# We could simply call `wrap` without tracer_provider, but then it would be bound to our global state
+Braintrust::Trace::OpenAI.wrap(client, tracer_provider: tracer_provider_b)
+
+puts "\nProject B: Describe Image"
+puts "=" * 50
+
+# chat completion should automatically nest
+root_span_b = nil
+tracer_b.in_span("vision") do |span|
+  root_span_b = span
+  # Example 1: Vision - Image Understanding
+  puts "\n Vision (Image Understanding)"
+  puts "-" * 50
+
+  input = "Tell me about this landmark."
+  tracer_b.in_span("example-vision") do |nested|
+    response = client.chat.completions.create(
+      model: model1,
+      messages: [
+        {
+          role: "user",
+          content: [
+            {type: "text", text: input},
+            {
+              type: "image_url",
+              image_url: {
+                url: url
+              }
+            }
+          ]
+        }
+      ],
+      max_tokens: 100
+    )
+
+    # Using Braintrust native span attributes
+    # For comparisons with OTEL GenAI semantic convention properties,
+    # see https://www.braintrust.dev/docs/integrations/sdk-integrations/opentelemetry#manual-tracing
+    nested.set_attribute("braintrust.span_attributes.type", "llm")
+    nested.set_attribute("metadata.model", model1)
+    nested.set_attribute("braintrust.input", input)
+    nested.set_attribute("braintrust.output", response.choices[0].message.content.to_s)
+
+    puts "✓ Vision response: #{response.choices[0].message.content[0..100]}..."
+    puts "  Tokens: #{response.usage.total_tokens}"
+  rescue OpenAI::Errors::BadRequestError => e
+    puts "⊘ Skipped - Image URL error (#{e.message.split("\n").first[0..80]}...)"
+  rescue => e
+    puts "⊘ Error: #{e.class}"
+  end
+end
+
+puts "\n✓ Vision example completed"
+puts "\n✓ View Project B trace in Braintrust:"
+puts "  #{Braintrust::Trace.permalink(root_span_b)}"
+
+# Shutdown both tracer providers to flush spans
+tracer_provider_a.shutdown
+tracer_provider_b.shutdown

gemfiles/ruby_llm_openai.gemfile

@@ -0,0 +1,9 @@
+# This file was generated by Appraisal
+
+source "https://rubygems.org"
+
+gem "minitest-reporters", "~> 1.6"
+gem "openai", ">= 0.34"
+gem "ruby_llm", ">= 1.9"
+
+gemspec path: "../"