feat: add Gemini embed_content tracking#498
Merged
carlos-marchal-ph merged 4 commits intomasterfrom Apr 10, 2026
Merged
Conversation
Contributor
posthog-python Compliance ReportDate: 2026-04-10 15:06:01 UTC ✅ All Tests Passed!0/0 tests passed |
Contributor
Prompt To Fix All With AIThis is a comment left during a code review.
Path: posthog/ai/gemini/gemini.py
Line: 28
Comment:
**Duplicate import from `posthog.ai.utils`**
`with_privacy_mode` is imported in a separate statement when an import for the same module already exists on lines 17–21. This violates DRY and makes the import block harder to scan.
```suggestion
from posthog.ai.utils import (
call_llm_and_track_usage,
capture_streaming_event,
merge_usage_stats,
with_privacy_mode,
)
```
The standalone `from posthog.ai.utils import with_privacy_mode` line (28) should then be removed. The same issue exists in `gemini_async.py`.
How can I resolve this? If you propose a fix, please make it concise.
---
This is a comment left during a code review.
Path: posthog/test/ai/gemini/test_gemini.py
Line: 1196-1351
Comment:
**Prefer parameterised tests**
The 8 new `embed_content` unit tests (and their 8 async mirrors in `test_gemini_async.py`) are written as independent functions, but several of them vary only in one dimension (input type, presence of token stats, privacy flag, etc.) and share nearly identical assertion logic. Per the project's testing convention, these should use `pytest.mark.parametrize`.
For example, `test_embed_content_basic` and `test_embed_content_without_token_counts` both call `embed_content` with a single string and verify the captured event — the only difference is which fixture is used and the expected `$ai_input_tokens` value. Similarly, `test_embed_content_privacy_mode` and `test_embed_content_no_distinct_id` each test a single property on the same captured event and could be collapsed into a single parameterised test.
Additionally, the `mock_embed_content_response` and `mock_embed_content_response_with_stats` fixtures are duplicated verbatim in both test files; moving them to a shared `conftest.py` would remove the redundancy.
How can I resolve this? If you propose a fix, please make it concise.Reviews (1): Last reviewed commit: "chore: add sampo changeset for gemini em..." | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| extract_gemini_content_from_chunk, | ||
| format_gemini_streaming_output, | ||
| ) | ||
| from posthog.ai.utils import with_privacy_mode |
Contributor
There was a problem hiding this comment.
posthog.ai.utils
with_privacy_mode is imported in a separate statement when an import for the same module already exists on lines 17–21. This violates DRY and makes the import block harder to scan.
Suggested change
| from posthog.ai.utils import with_privacy_mode | |
| from posthog.ai.utils import ( | |
| call_llm_and_track_usage, | |
| capture_streaming_event, | |
| merge_usage_stats, | |
| with_privacy_mode, | |
| ) |
The standalone from posthog.ai.utils import with_privacy_mode line (28) should then be removed. The same issue exists in gemini_async.py.
Prompt To Fix With AI
This is a comment left during a code review.
Path: posthog/ai/gemini/gemini.py
Line: 28
Comment:
**Duplicate import from `posthog.ai.utils`**
`with_privacy_mode` is imported in a separate statement when an import for the same module already exists on lines 17–21. This violates DRY and makes the import block harder to scan.
```suggestion
from posthog.ai.utils import (
call_llm_and_track_usage,
capture_streaming_event,
merge_usage_stats,
with_privacy_mode,
)
```
The standalone `from posthog.ai.utils import with_privacy_mode` line (28) should then be removed. The same issue exists in `gemini_async.py`.
How can I resolve this? If you propose a fix, please make it concise.Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
Comment on lines
+1196
to
+1351
| def test_embed_content_basic(mock_client, mock_google_genai_client, mock_embed_content_response): | ||
| mock_google_genai_client.models.embed_content.return_value = mock_embed_content_response | ||
|
|
||
| client = Client(api_key="test-key", posthog_client=mock_client) | ||
| response = client.models.embed_content( | ||
| model="gemini-embedding-001", | ||
| contents="Hello world", | ||
| posthog_distinct_id="test-id", | ||
| posthog_properties={"foo": "bar"}, | ||
| ) | ||
|
|
||
| assert response == mock_embed_content_response | ||
| assert mock_client.capture.call_count == 1 | ||
|
|
||
| call_args = mock_client.capture.call_args[1] | ||
| props = call_args["properties"] | ||
|
|
||
| assert call_args["distinct_id"] == "test-id" | ||
| assert call_args["event"] == "$ai_embedding" | ||
| assert props["$ai_provider"] == "gemini" | ||
| assert props["$ai_model"] == "gemini-embedding-001" | ||
| assert props["$ai_input"] == "Hello world" | ||
| assert props["$ai_http_status"] == 200 | ||
| assert isinstance(props["$ai_latency"], float) | ||
| assert "$ai_trace_id" in props | ||
| assert props["$ai_base_url"] == "https://generativelanguage.googleapis.com" | ||
| assert props["foo"] == "bar" | ||
|
|
||
| # Verify the underlying call received the right args | ||
| mock_google_genai_client.models.embed_content.assert_called_once_with( | ||
| model="gemini-embedding-001", contents="Hello world" | ||
| ) | ||
|
|
||
|
|
||
| def test_embed_content_with_token_counts(mock_client, mock_google_genai_client, mock_embed_content_response_with_stats): | ||
| mock_google_genai_client.models.embed_content.return_value = mock_embed_content_response_with_stats | ||
|
|
||
| client = Client(api_key="test-key", posthog_client=mock_client) | ||
| client.models.embed_content( | ||
| model="gemini-embedding-001", | ||
| contents=["Hello", "World"], | ||
| posthog_distinct_id="test-id", | ||
| ) | ||
|
|
||
| props = mock_client.capture.call_args[1]["properties"] | ||
| assert props["$ai_input_tokens"] == 13 # 5 + 8 | ||
|
|
||
|
|
||
| def test_embed_content_without_token_counts(mock_client, mock_google_genai_client, mock_embed_content_response): | ||
| mock_google_genai_client.models.embed_content.return_value = mock_embed_content_response | ||
|
|
||
| client = Client(api_key="test-key", posthog_client=mock_client) | ||
| client.models.embed_content( | ||
| model="gemini-embedding-001", | ||
| contents="Hello", | ||
| posthog_distinct_id="test-id", | ||
| ) | ||
|
|
||
| props = mock_client.capture.call_args[1]["properties"] | ||
| assert props["$ai_input_tokens"] == 0 | ||
|
|
||
|
|
||
| def test_embed_content_privacy_mode(mock_client, mock_google_genai_client, mock_embed_content_response): | ||
| mock_google_genai_client.models.embed_content.return_value = mock_embed_content_response | ||
|
|
||
| client = Client(api_key="test-key", posthog_client=mock_client) | ||
| client.models.embed_content( | ||
| model="gemini-embedding-001", | ||
| contents="Secret text", | ||
| posthog_distinct_id="test-id", | ||
| posthog_privacy_mode=True, | ||
| ) | ||
|
|
||
| props = mock_client.capture.call_args[1]["properties"] | ||
| assert props["$ai_input"] is None | ||
|
|
||
|
|
||
| def test_embed_content_no_distinct_id(mock_client, mock_google_genai_client, mock_embed_content_response): | ||
| mock_google_genai_client.models.embed_content.return_value = mock_embed_content_response | ||
|
|
||
| client = Client(api_key="test-key", posthog_client=mock_client) | ||
| client.models.embed_content( | ||
| model="gemini-embedding-001", | ||
| contents="Hello", | ||
| ) | ||
|
|
||
| call_args = mock_client.capture.call_args[1] | ||
| props = call_args["properties"] | ||
|
|
||
| # Should fall back to trace_id as distinct_id | ||
| assert call_args["distinct_id"] == props["$ai_trace_id"] | ||
| assert props["$process_person_profile"] is False | ||
|
|
||
|
|
||
| def test_embed_content_default_params(mock_client, mock_google_genai_client, mock_embed_content_response): | ||
| mock_google_genai_client.models.embed_content.return_value = mock_embed_content_response | ||
|
|
||
| client = Client( | ||
| api_key="test-key", | ||
| posthog_client=mock_client, | ||
| posthog_distinct_id="default-id", | ||
| posthog_properties={"team": "ai"}, | ||
| posthog_groups={"company": "test-co"}, | ||
| ) | ||
| client.models.embed_content( | ||
| model="gemini-embedding-001", | ||
| contents="Hello", | ||
| posthog_properties={"extra": "prop"}, | ||
| ) | ||
|
|
||
| call_args = mock_client.capture.call_args[1] | ||
| props = call_args["properties"] | ||
|
|
||
| assert call_args["distinct_id"] == "default-id" | ||
| assert props["team"] == "ai" | ||
| assert props["extra"] == "prop" | ||
| assert call_args["groups"] == {"company": "test-co"} | ||
|
|
||
|
|
||
| def test_embed_content_error_handling(mock_client, mock_google_genai_client): | ||
| mock_google_genai_client.models.embed_content.side_effect = Exception("API error") | ||
|
|
||
| client = Client(api_key="test-key", posthog_client=mock_client) | ||
|
|
||
| with pytest.raises(Exception, match="API error"): | ||
| client.models.embed_content( | ||
| model="gemini-embedding-001", | ||
| contents="Hello", | ||
| posthog_distinct_id="test-id", | ||
| ) | ||
|
|
||
| # Event should still be captured | ||
| assert mock_client.capture.call_count == 1 | ||
| props = mock_client.capture.call_args[1]["properties"] | ||
| assert props["$ai_is_error"] is True | ||
| assert props["$ai_error"] == "API error" | ||
| assert props["$ai_http_status"] == 0 | ||
|
|
||
|
|
||
| def test_embed_content_kwargs_passthrough(mock_client, mock_google_genai_client, mock_embed_content_response): | ||
| mock_google_genai_client.models.embed_content.return_value = mock_embed_content_response | ||
|
|
||
| client = Client(api_key="test-key", posthog_client=mock_client) | ||
| client.models.embed_content( | ||
| model="gemini-embedding-001", | ||
| contents="Hello", | ||
| posthog_distinct_id="test-id", | ||
| config={"output_dimensionality": 64}, | ||
| ) | ||
|
|
||
| mock_google_genai_client.models.embed_content.assert_called_once_with( | ||
| model="gemini-embedding-001", | ||
| contents="Hello", | ||
| config={"output_dimensionality": 64}, | ||
| ) | ||
|
|
Contributor
There was a problem hiding this comment.
The 8 new embed_content unit tests (and their 8 async mirrors in test_gemini_async.py) are written as independent functions, but several of them vary only in one dimension (input type, presence of token stats, privacy flag, etc.) and share nearly identical assertion logic. Per the project's testing convention, these should use pytest.mark.parametrize.
For example, test_embed_content_basic and test_embed_content_without_token_counts both call embed_content with a single string and verify the captured event — the only difference is which fixture is used and the expected $ai_input_tokens value. Similarly, test_embed_content_privacy_mode and test_embed_content_no_distinct_id each test a single property on the same captured event and could be collapsed into a single parameterised test.
Additionally, the mock_embed_content_response and mock_embed_content_response_with_stats fixtures are duplicated verbatim in both test files; moving them to a shared conftest.py would remove the redundancy.
Prompt To Fix With AI
This is a comment left during a code review.
Path: posthog/test/ai/gemini/test_gemini.py
Line: 1196-1351
Comment:
**Prefer parameterised tests**
The 8 new `embed_content` unit tests (and their 8 async mirrors in `test_gemini_async.py`) are written as independent functions, but several of them vary only in one dimension (input type, presence of token stats, privacy flag, etc.) and share nearly identical assertion logic. Per the project's testing convention, these should use `pytest.mark.parametrize`.
For example, `test_embed_content_basic` and `test_embed_content_without_token_counts` both call `embed_content` with a single string and verify the captured event — the only difference is which fixture is used and the expected `$ai_input_tokens` value. Similarly, `test_embed_content_privacy_mode` and `test_embed_content_no_distinct_id` each test a single property on the same captured event and could be collapsed into a single parameterised test.
Additionally, the `mock_embed_content_response` and `mock_embed_content_response_with_stats` fixtures are duplicated verbatim in both test files; moving them to a shared `conftest.py` would remove the redundancy.
How can I resolve this? If you propose a fix, please make it concise.Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
Contributor
Author
|
This PR was implemented to address a community request for tracking Gemini embedding calls. JS SDK companion PR for feature parity: PostHog/posthog-js#3369 |
7 tasks
andrewm4894
approved these changes
Apr 10, 2026
…mbed-content # Conflicts: # posthog/test/ai/gemini/test_gemini.py
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Changes
Adds
embed_contentwrapping to the Gemini client (both sync and async), capturing$ai_embeddingevents in PostHog. This addresses a community request for tracking Gemini embedding calls.embed_content()toModelsandAsyncModelsclasses, matching the nativeclient.models.embed_content()API$ai_embeddingevents with provider, model, input, latency, and token counts (Vertex AI)How to test
pytest posthog/test/ai/gemini/ -v— 20 new tests (8 sync + 8 async mock-based, 2 sync + 2 async integration)GOOGLE_API_KEY=$GEMINI_API_KEYand hit the real Gemini embedding API