- Support
- Integrations
- Generative AI
Generative AI
Oracle Cloud integration · 14 node(s).
00Overview
Call Oracle Cloud's Generative AI models straight from a flow — chat with a model, generate or summarize text, turn text into vector embeddings for search, and rerank documents by relevance to a query. Manage the hosting side too: browse the available models and dedicated AI clusters, and create, inspect, move and tear down the endpoints that serve a model for inference. Every action runs against a single OCI region and compartment, so model output flows straight into the rest of your automation.
Every field below is exactly what you see in the Flomation editor. Fields marked ● live picker let you choose from a list pulled live from your account — no IDs to look up.
01Connecting Generative AI
- Decide how the node signs its calls with the Authentication dropdown: Connect Oracle Cloud uses a managed connection — just pick your account in the Oracle Cloud connection field and skip the manual key steps below — while API signing key (advanced) lets you paste raw OCI signing credentials.
- For the advanced method, sign in to the OCI Console, open the Profile menu (top-right) → User settings, then Resources → API keys and choose Add API key. Generate a new key pair and download the private key.
- OCI now shows a Configuration file preview — copy its values into the matching node fields: Tenancy OCID, User OCID, Key Fingerprint, and Region (use a region where Generative AI is available, e.g.
uk-london-1). - In Flomation, store the downloaded private key — the full PEM, including its
BEGIN/ENDlines — as an environment secret (e.g.generativeai_secret) and pick it in the node's Private Key (PEM) field; if the key is encrypted, add its passphrase in Private Key Passphrase. - Set Compartment OCID on every node — find it under Identity & Security → Compartments in the OCI Console. It scopes both the inference calls (chat, generate, summarize, embed, rerank) and the management calls (models, endpoints, clusters).
| Field | Type | Details | |
|---|---|---|---|
| Authentication | string | Connect Oracle Cloud, API signing key (advanced) | |
| Oracle Cloud connection | credential | Pick a connected Oracle Cloud account | |
| Region | string | e.g. uk-london-1 | |
| Private Key (PEM) | secret | The API signing private key — full PEM, incl. BEGIN/END lines | |
| Private Key Passphrase | secret | Only if the key is encrypted (optional) | |
| Tenancy OCID | string | ocid1.tenancy.oc1..aaaa… | |
| User OCID | string | ocid1.user.oc1..aaaa… | |
| Key Fingerprint | string | aa:bb:cc:… fingerprint of the uploaded API key |
Pick an Environment on your flow (Flow Settings → Environment) so the secret resolves. Secret fields never show the value — they reference ${secrets.your_secret}.
02Chat
OCI Generative AI: Chat
oracle/generativeai/chat · Action
Send a message to an on-demand Cohere chat model and return its reply.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Model OCID | string | Required | ocid1.generativeaimodel.oc1..aaaa… — an on-demand Cohere chat model |
| Message | text | Required | The text you want the model to respond to |
Returns: tool_result, response, success, error
03Dedicated
OCI Generative AI: List Dedicated AI Clusters
oracle/generativeai/dedicated_ai_cluster_list · Action
List the dedicated AI clusters in a compartment. Optionally filter by exact display name or lifecycle state, and cap the page size. Walks pagination up to a safe cap.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… (use the tenancy OCID for the root) |
| Display Name Filter | string | Only clusters with this exact name (optional) | |
| Lifecycle State | string | Filter by lifecycle state (optional) — choices: Creating, Active, Updating, Deleting, Deleted, Failed, Needs Attention | |
| Page Size | string | Max results per page (optional) |
Returns: tool_result, clusters, count, truncated, success, error
04Embed
OCI Generative AI: Embed Text
oracle/generativeai/embed_text · Action
Turn one or more input strings into vector embeddings with an on-demand OCI Generative AI embedding model.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Embedding Model OCID / Name | string | Required | e.g. cohere.embed-english-v3.0 or an ocid1.generativeaimodel… OCID |
| Inputs (comma-separated) | text | Required | One or more strings to embed, comma-separated. Each up to 512 tokens. |
| Truncate | string | How to trim inputs longer than the model limit (default NONE) — choices: None (error if too long), Start, End |
Returns: tool_result, embeddings, count, dimension, model_id, id, success, error
05Endpoint
OCI Generative AI: Change Endpoint Compartment
oracle/generativeai/endpoint_change_compartment · Action
Move a Generative AI endpoint into a different compartment — the endpoint keeps its OCID, only its compartment placement changes.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Endpoint OCID | string | Required | ocid1.generativeaiendpoint.oc1..aaaa… (the endpoint to move) |
| Destination Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… (where to move the endpoint) |
Returns: tool_result, id, destination_compartment_id, success, error
OCI Generative AI: Create Endpoint
oracle/generativeai/endpoint_create · Action
Create an endpoint that hosts a model on a dedicated AI cluster for inference. Returns a work-request id — poll Get Endpoint until ACTIVE.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Model OCID | string | Required | ocid1.generativeaimodel.oc1..aaaa… — the model to host |
| Dedicated AI Cluster OCID | string | Required | ocid1.generativeaidedicatedaicluster.oc1..aaaa… — where the model is deployed |
| Display Name | string | A name for the endpoint (optional) |
Returns: tool_result, work_request_id, success, error
OCI Generative AI: Delete Endpoint
oracle/generativeai/endpoint_delete · Action
Delete a Generative AI endpoint by its OCID — it tears down the hosting for its model. Returns a work-request id to track the async teardown.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Endpoint OCID | string | Required | ocid1.generativeaiendpoint.oc1..aaaa… of the endpoint to delete |
Returns: tool_result, id, work_request_id, success, error
OCI Generative AI: Get Endpoint
oracle/generativeai/endpoint_get · Action
Fetch a single Generative AI endpoint by its OCID — its model, dedicated AI cluster, description and lifecycle state.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Endpoint OCID | string | Required | ocid1.generativeaiendpoint.oc1..aaaa… |
Returns: tool_result, endpoint, id, lifecycle_state, success, error
OCI Generative AI: List Endpoints
oracle/generativeai/endpoint_list · Action
List the Generative AI endpoints in a compartment. Optionally filter by exact display name or lifecycle state, and cap the page size. Walks pagination up to a safe cap.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… (use the tenancy OCID for the root) |
| Display Name Filter | string | Only endpoints with this exact name (optional) | |
| Lifecycle State | string | Filter by state (optional) — choices: Active, Creating, Updating, Deleting, Deleted, Failed | |
| Page Size Limit | string | Max results per page (optional) |
Returns: tool_result, endpoints, count, truncated, success, error
OCI Generative AI: Update Endpoint
oracle/generativeai/endpoint_update · Action
Partially update a Generative AI endpoint — change only the display name or description you supply; blank fields are left unchanged. Asynchronous: returns a work-request id, poll Get Endpoint until ACTIVE.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Endpoint OCID | string | Required | ocid1.generativeaiendpoint.oc1..aaaa… — the endpoint to update |
| Display Name | string | New name (leave blank to keep unchanged) | |
| Description | string | New description (leave blank to keep unchanged) |
Returns: tool_result, endpoint, id, work_request_id, success, error
06Generate
OCI Generative AI: Generate Text
oracle/generativeai/generate_text · Action
Run a single text-generation (completion) request against an on-demand Generative AI model and return the generated text.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Model ID | string | Required | On-demand model OCID or name, e.g. cohere.command-r-08-2024 |
| Prompt | text | Required | The prompt to complete |
Returns: tool_result, response, generations, model_id, model_version, success, error
07Model
OCI Generative AI: Get Model
oracle/generativeai/model_get · Action
Fetch a single Generative AI model by its OCID — its vendor, version, capabilities, base model and lifecycle state.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Model OCID | string | Required | ocid1.generativeaimodel.oc1..aaaa… |
Returns: tool_result, model, id, lifecycle_state, success, error
OCI Generative AI: List Models
oracle/generativeai/model_list · Action
List the Generative AI models in a compartment. Optionally filter by capability, vendor or lifecycle state. Walks pagination up to a safe cap.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… (use the tenancy OCID for the root) |
| Capability Filter | string | Only models with this capability (optional) — choices: Chat, Text Generation, Text Summarization, Text Embeddings, Text Rerank, Fine-Tune | |
| Vendor Filter | string | Only models from this exact vendor, e.g. cohere or meta (optional) | |
| Lifecycle State | string | Only models in this state (optional) — choices: Active, Creating, Deleting, Deleted, Failed | |
| Page Size | string | Max results per page (optional) |
Returns: tool_result, models, count, truncated, success, error
08Rerank
OCI Generative AI: Rerank Text
oracle/generativeai/rerank_text · Action
Rerank a list of documents by relevance to a query using an OCI Generative AI rerank model, returning each document's original index and relevance score.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Rerank Model | string | Required | OCID or name of a rerank model (e.g. cohere.rerank-multilingual-v3.1) |
| Query | text | Required | The search query to rank the documents against |
| Documents | text | Required | One document per line — each line is ranked against the query |
| Top N | string | How many of the most relevant documents to return (optional, defaults to all) |
Returns: tool_result, ranks, success, error
09Summarize
OCI Generative AI: Summarize Text
oracle/generativeai/summarize_text · Action
Summarize a block of text with an on-demand OCI Generative AI model and return the summary.
| Field | Type | Details | |
|---|---|---|---|
| Compartment OCID | string | Required | ocid1.compartment.oc1..aaaa… |
| Model OCID | string | Required | ocid1.generativeaimodel.oc1..aaaa… (a model with the TEXT_SUMMARIZATION capability) |
| Input Text | string | Required | The text to summarize |
Returns: tool_result, summary, model_id, success, error
10Notes & Limitations
Behaviours and constraints worth knowing before you build with these nodes.
- Chat and Generate Text read the reply only from Cohere-family models; point either action at a model from another vendor and the call still runs but the returned text comes back empty.
- The inference actions — chat, generate, summarize, embed and rerank — run in on-demand serving mode, so give them a pretrained model's OCID or name rather than a hosted endpoint's OCID.
- Embed Text splits the Inputs field on commas, so a single value that itself contains a comma is embedded as several separate inputs.
- Rerank Text treats each line of the Documents field as its own document, so keep every document on a single line.
- The list actions stop after 25 pages of results, so in large compartments narrow the output with the available display-name, state, vendor, or capability filters rather than relying on a complete listing.