Google Gemini
Put Google Gemini to work on your own records, inside your rules, with a person approving what it changes.
23 ways to read, 9 ways to write, each write switched on by you.
Quality checks
Built and tested with example responses.
- Ready to connect: doneEvery listed action is built in.
- Built and tested with example responses: doneEach action is built from the provider’s published API and tested against example responses, errors, rate limits and paging.
Connected, approved, proven.
The things teams do most with Google Gemini, an example flow with a person in the loop, and what Nexra keeps as proof.
Top actions
- Create batchApproval by rule
- Create cached contentApproval by rule
- List batchesRead
Example flow
- Google Gemini · List batchesDone
- Google Gemini · Create batchNeeds approval
- Nexra · Record the decision and result in the evidence chainProven
Example workflow. Routine runs go through on their own. Only a change your rules flag, such as one over a limit you set, waits for the right person, who sees the exact change and the rule before it reaches Google Gemini.
What gets logged
- Whether it ran on its own or who approved it, and under which rule
- A fingerprint of the exact request sent to Google Gemini
- Google Gemini's reply: what it created or changed
- One entry in the evidence chain, linked to the one before
Try this in the sandbox
Run the whole flow with test data in your sandbox, then connect a Google test account before you go live.
For developers
Technical detail
Gemini generation, schema JSON, embeddings and token counts, plus context caches, files, Batch Mode (generation and embeddings), operations, tuned models and File Search stores.
Google API docs- Source actions
- 23
- Destination actions
- 9
- Data types
- 19
What you can map
Reads are always available. Writes are off until you switch them on for a connection, and governed writes follow your policies.
Batches3
- Create batchDestinationNot retried automaticallyApproval before production
Enqueues generateContent requests in Batch Mode at the batch discount, from an uploaded JSONL file or inline requests.
- Get batchSource
State, request counts and results file of one batch.
- List batchesSource
Batch Mode generation jobs with state and request counts.
Cached contents4
- Create cached contentDestinationNot retried automaticallyApproval before production
Caches a system instruction and reusable text for a model so later requests can reference it. Cached tokens are billed per hour of storage.
- Get cached contentSource
Metadata for one context cache.
- List cached contentsSource
Context caches in the project with model, token count and expiry (no contents).
- Extend cached contentDestinationProvider duplicate protectionApproval before production
Changes a cache's expiry; only the expiration can be updated.
Content1
- Generate contentDestinationNot retried automaticallyApproval before production
Generates a reply with generateContent. Supply a JSON schema to receive schema-constrained JSON.
Embedding batch jobs1
- Create embedding batchDestinationNot retried automaticallyApproval before production
Enqueues embedContent requests in Batch Mode at the batch discount, from an uploaded JSONL file or inline requests.
Embedding batches1
- Embed texts in batchDestinationNot retried automaticallyApproval before production
Creates embeddings for several texts with batchEmbedContents; one record per text.
Embeddings1
- Embed textDestinationNot retried automaticallyApproval before production
Creates one embedding with embedContent.
File search store documents2
- Get file search store documentSource
One document in a File Search store.
- List file search store documentsSource
Documents in a File Search store with state, size and custom metadata.
File search store imports1
- Import file into file search storeDestinationNot retried automaticallyApproval before production
Imports an uploaded file into a File Search store; Google chunks and embeds it (embeddings are billed at indexing). Returns a long-running operation.
File search store operations1
- Get file search store operationSource
Progress of a File Search import operation.
File search store upload operations1
- Get file search upload operationSource
Status of a direct upload into a File Search store.
File search stores3
- Create file search storeDestinationNot retried automaticallyApproval before production
Creates an empty File Search store for retrieval-augmented generation.
- Get file search storeSource
One File Search store.
- List file search storesSource
File Search stores with document counts and size.
Files2
- Get file metadataSource
Processing state, URI and expiry for one file.
- List filesSource
Files uploaded to the project (metadata only; files expire after 48 hours).
Generated files1
- List generated filesSource
Files generated by the API (for example batch outputs) with state and MIME type.
Model operations2
- Get model operationSource
Status of one long-running model operation (the generated result is not mapped).
- List model operationsSource
Long-running operations started against a model (for example video generation), with status only.
Models2
- Get modelSource
Metadata for one Gemini model.
- List modelsSource
Gemini models available to the key with token limits and supported methods.
Token count1
- Count tokensSource
Counts the tokens a prompt uses for a model (countTokens).
Tuned model operations2
- Get tuned model operationSource
Status of one tuning operation.
- List tuned model operationsSource
Long-running tuning operations for one tuned model, with status.
Tuned model permissions1
- List tuned model permissionsSource
Who can use a tuned model, as grantee type and role (grantee email addresses are not mapped).
Tuned models2
- Get tuned modelSource
One tuned model with tuning hyperparameters and timings.
- List tuned modelsSource
Tuned models owned by the project with base model and state.
Access and limits
- Authentication
- Gemini API key (x-goog-api-key header)
- Plans and access
- Gemini API keys from Google AI Studio on the free or paid tier. Free-tier usage may be used by Google to improve products and has lower rate limits; use a paid-tier project for business data.
- Test environment
- The free tier can be used for testing with reduced limits.
- Rate limits
- Per-project requests and tokens per minute and requests per day by model and tier; HTTP 429 when exceeded.
- Pagination
- pageToken / nextPageToken cursors for models, caches, files, batches, tuned models, File Search stores and documents.
- Provider events
- Gemini API does not deliver webhooks. Flows run on demand, on a schedule, or when a scheduled check finds new or changed records.
Before you build a flow
- Uses the v1beta API. Streaming, tools, media inputs in generation and the Interactions API are not exposed.
- File uploads (resumable/multipart media upload) and file downloads are not implemented; File Search imports and batch inputs reference files uploaded outside Nexra.
- Context cache creation, Batch Mode creation and File Search imports are billable and never retried automatically. Caches can be extended but not deleted from Nexra.
- Batch results in a responses file must be downloaded outside Nexra; inline batch responses are not mapped. Cancelling batches is not implemented.
- Embedding batches (asyncBatchEmbedContent) are billable and never retried; they appear in batches.list and batches.get. Updating, cancelling or deleting batches is not implemented.
- Long-running model and tuning operations, generated files and upload operations are read as status only; generated media, results and tuned model permission grantee email addresses are never mapped.
- Live API auth tokens, the Interactions API environments and legacy PaLM methods (generateText, embedText, generateMessage, generateAnswer) are not implemented.
- Tuned models are read only: Google currently offers no Gemini API model that supports tuning, so creating tuned models is not implemented.
- Corpora (Semantic Retriever) are deprecated in favour of File Search stores and are not implemented. Deleting caches, files, stores and documents is intentionally excluded.
- Structured output uses generationConfig.responseMimeType with responseJsonSchema. Google's reference now also lists a newer responseFormat field; it is not used yet.
- Embedding options are sent in embedContentConfig; the older top-level taskType, title and outputDimensionality fields are deprecated.
- Token counting is a read; Google's reference does not state its price, so verify on the account's billing page.
- Generation and embeddings are billed; they must be enabled per connection and are never retried automatically.
Developer reference
- Create the Google credentials described above with the narrowest permissions your flows need.
- In Nexra, open Connect, then Connections, then New connection, choose Google Gemini, enter its settings and credentials, and tick only the write actions you need. Connection guide
- Test the connection, then use its actions as flow sources and destinations. Flow guide
Machine-readable detail: GET /api/public/connect/connectors/google-gemini and the Connections API.