references/google.md
AI agents: this is one page from PostHog's docs. Full index of Markdown docs for LLMs: https://posthog.com/llms.txt
Google AI Observability installation
Canonical URL: https://posthog.com/docs/ai-observability/installation/google
Cloud region: This page uses US Cloud hosts. For an EU Cloud project, use
https://eu.i.posthog.cominstead ofhttps://us.i.posthog.com(ingestion host), and usehttps://eu.posthog.cominstead ofhttps://app.posthog.com(app host for links).
On this page
- Install dependencies
- Configure PostHog
- Call Google Gen AI LLMs
- Capture tool calls as spans
- Capture embeddings
- Verify traces and generations
- Next steps
Let AI instrument your LLM calls for you
Skip the manual setup — run this in your project and the wizard installs the SDK and wires up AI Observability for you.
npx @posthog/wizard ai-observability
PostHog Wizard hedgehog
Let AI instrument your LLM calls for you
-
1
Install dependencies
Required
Full working examples
See the complete Node.js and Python examples on GitHub.
Install the PostHog SDK and the Google Gen AI SDK.
Python
pip install posthog google-genaiNode
npm install @posthog/ai posthog-node @google/genaiGo
PostHog doesn't have a Go wrapper for this SDK yet. Use the [OpenTelemetry integration](https://posthog.com/docs/ai-observability/installation/opentelemetry) instead. -
2
Configure PostHog
Required
Create a PostHog client, then swap in PostHog's Google Gen AI wrapper.
Python
from posthog import Posthog from posthog.ai.gemini import Client import time, uuid from google.genai import types posthog = Posthog("<ph_project_token>", host="https://us.i.posthog.com") client = Client( api_key="your_gemini_api_key", posthog_client=posthog, )Node
import { GoogleGenAI } from '@posthog/ai/gemini' import { PostHog } from 'posthog-node' const posthog = new PostHog('<ph_project_token>', { host: 'https://us.i.posthog.com' }) const client = new GoogleGenAI({ apiKey: 'your_gemini_api_key', posthog, })To call Gemini through Vertex AI, pass your Google Cloud project and location instead of an API key. The Google Gen AI client then authenticates with your Google Cloud credentials.
Python
client = Client( vertexai=True, project="your_gcp_project_id", location="us-central1", posthog_client=posthog, )Node
const client = new GoogleGenAI({ vertexai: true, project: 'your_gcp_project_id', location: 'us-central1', posthog, }) -
3
Call Google Gen AI LLMs
Required
When you use the wrapped client to call Gemini, PostHog automatically captures an
$ai_generationevent.Audio output usage is captured separately from text output when Gemini reports the modality breakdown. Dedicated standard Gemini Developer API audio pricing is available for
gemini-2.5-pro-preview-ttsandgemini-3.1-flash-tts-preview. Other audio models, including newer TTS and Live models, need their own catalog rates or custom pricing; capturing their usage alone does not establish accurate audio pricing. Unlisted providers use the catalog's default estimate, rather than a verified provider-specific rate.Python
trace_id = str(uuid.uuid4()) response = client.models.generate_content( model="gemini-2.5-flash", contents=[{"role": "user", "parts": [{"text": "What's the weather in Paris?"}]}], config=types.GenerateContentConfig(tools=tools), posthog_distinct_id="user_123", posthog_trace_id=trace_id, posthog_properties={ "$ai_session_id": "conversation-abc", }, )Node
const traceId = crypto.randomUUID() const response = await client.models.generateContent({ model: 'gemini-2.5-flash', contents: "What's the weather in Paris?", config: { tools }, posthogDistinctId: 'user_123', posthogTraceId: traceId, posthogProperties: { $ai_session_id: 'conversation-abc', }, })Calls that share a
posthog_trace_id(posthogTraceIdin Node) appear as one trace. If you omit it, the wrapper creates a new trace ID for each call, so each call appears as its own trace. To see a multi-step task as one trace, create one trace ID for the task and pass it to every call in that task.Note: If you want to capture LLM events anonymously, omit
posthog_distinct_idfrom the call. See our docs on anonymous vs identified events to learn more.You can expect captured
$ai_generationevents to have the following properties:| Property | Description | | --- | --- | |
$ai_model| The specific model, likegpt-5-miniorclaude-4-sonnet| |$ai_latency| The latency of the LLM call in seconds | |$ai_time_to_first_token| Time to first token in seconds (streaming only) | |$ai_tools| Tools and functions available to the LLM | |$ai_input| List of messages sent to the LLM | |$ai_input_tokens| The number of tokens in the input (often found in response.usage) | |$ai_output_choices| List of response choices from the LLM | |$ai_output_tokens| The number of tokens in the output (often found inresponse.usage) | |$ai_total_cost_usd| The total cost in USD (input + output) | | [...] | See full list of properties | -
4
Capture tool calls as spans
Optional
For standard responses, the posthog client captures it as a generation. For all tool calls, you must manually capture them as
$ai_spanevents.Python
for call in response.function_calls or []: start = time.time() result = run_tool(call.name, call.args) posthog.capture( distinct_id="user_123", event="$ai_span", properties={ "$ai_trace_id": trace_id, "$ai_session_id": "conversation-abc", "$ai_span_id": str(uuid.uuid4()), "$ai_span_name": call.name, "$ai_input_state": call.args, "$ai_output_state": result, "$ai_latency": time.time() - start, }, )Node
for (const call of response.functionCalls ?? []) { const start = Date.now() const result = await runTool(call.name, call.args) posthog.capture({ distinctId: 'user_123', event: '$ai_span', properties: { $ai_trace_id: traceId, $ai_session_id: 'conversation-abc', $ai_span_id: crypto.randomUUID(), $ai_span_name: call.name, $ai_input_state: call.args, $ai_output_state: result, $ai_latency: (Date.now() - start) / 1000, }, }) }See spans for the full list of span properties.
-
5
Capture embeddings
Optional
PostHog can also capture embedding generations as
$ai_embeddingevents. The wrapped client captures these automatically when you use theembed_contentAPI:Python
response = client.models.embed_content( model="gemini-embedding-001", contents="The quick brown fox", )Node
const response = await client.models.embedContent({ model: 'gemini-embedding-001', contents: 'The quick brown fox', }) -
Verify traces and generations
Recommended
Confirm LLM events are being sent to PostHog
Let's make sure LLM events are being captured and sent to PostHog. Under AI Observability, you should see rows of data appear in the Traces and Generations tabs.
LLM generations in PostHog
-
6
Next steps
Recommended
Now that you're capturing AI conversations, continue with the resources below to learn what else AI Observability enables within the PostHog platform.
| Resource | Description | | --- | --- | | Basics | Learn the basics of how LLM calls become events in PostHog. | | Generations | Read about the
$ai_generationevent and its properties. | | Traces | Explore the trace hierarchy and how to use it to debug LLM calls. | | Spans | Review spans and their role in representing individual operations. | | Anaylze LLM performance | Learn how to create dashboards to analyze LLM performance. |