The PaLM 2 for Chat (chat-bison
) foundation model is a large language model (LLM) that excels at
language understanding, language generation, and conversations. This chat model
is fine-tuned to conduct natural multi-turn conversations, and is ideal for text
tasks about code that require back-and-forth interactions.
For text tasks that can be completed with one API response (without the need for continuous conversation), use the Text model.
To explore this model in the console, see the PaLM 2 for Chat model card in the
Model Garden.
Go to the Model Garden
Use cases
Customer Service: Instruct the model to respond as customer service agents that only talk about your company's product
Technical Support: Instruct the model to interact with customers as a call center agent with specific parameters about how to respond and what not to say
Personas and characters: Instruct the model to respond in the style of a specific person ("...in the style of Shakespeare")
Website companion: Create a conversational assistant for shopping, travel, and other use cases
For more information, see Design chat prompts.
HTTP request
POST https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/chat-bison:predict
For more information, see the predict
method.
Model versions
To use the latest model version,
specify the model name without a version number, for example chat-bison
.
To use a stable model version,
specify the model version number, for example chat-bison@002
. Each
stable version is available for six months after the release date of the
subsequent stable version.
The following table contains the available stable model versions:
chat-bison model | Release date | Discontinuation date |
---|---|---|
chat-bison@002 | December 6, 2023 | April 9, 2025 |
For more information, see Model versions and lifecycle.
Request body
{
"instances": [
{
"context": string,
"examples": [
{
"input": { "content": string },
"output": { "content": string }
}
],
"messages": [
{
"author": string,
"content": string,
}
],
}
],
"parameters": {
"temperature": number,
"maxOutputTokens": integer,
"topP": number,
"topK": integer,
"groundingConfig": string,
"stopSequences": [ string ],
"candidateCount": integer
"logprobs": integer,
"presencePenalty": float,
"frequencyPenalty": float,
"seed": integer
}
}
For chat API calls, the context
, examples
, and messages
combine to form
the prompt. The following table shows the parameters that you need to configure
for the Vertex AI PaLM API for text:
Parameter | Description | Acceptable values |
---|---|---|
(optional) |
Context shapes how the model responds throughout the conversation. For example, you can use context to specify words the model can or cannot use, topics to focus on or avoid, or the response format or style. | Text |
(optional) |
Examples for the model to learn how to respond to the conversation. | [{ "input": {"content": "provide content"}, "output": {"content": "provide content"} }] |
(required) |
Conversation history provided to the model in a structured alternate-author form. Messages appear in chronological order: oldest first, newest last. When the history of messages causes the input to exceed the maximum length, the oldest messages are removed until the entire prompt is within the allowed limit. | [{ "author": "user", "content": "user message" }] |
|
The temperature is used for sampling during response generation, which occurs when topP
and topK are applied. Temperature controls the degree of randomness in token selection.
Lower temperatures are good for prompts that require a less open-ended or creative response, while
higher temperatures can lead to more diverse or creative results. A temperature of 0
means that the highest probability tokens are always selected. In this case, responses for a given
prompt are mostly deterministic, but a small amount of variation is still possible.
If the model returns a response that's too generic, too short, or the model gives a fallback response, try increasing the temperature. |
|
|
Maximum number of tokens that can be generated in the response. A token is
approximately four characters. 100 tokens correspond to roughly 60-80 words.
Specify a lower value for shorter responses and a higher value for potentially longer responses. |
|
|
Top-K changes how the model selects tokens for output. A top-K of
1 means the next selected token is the most probable among all
tokens in the model's vocabulary (also called greedy decoding), while a top-K of
3 means that the next token is selected from among the three most
probable tokens by using temperature.
For each token selection step, the top-K tokens with the highest probabilities are sampled. Then tokens are further filtered based on top-P with the final token selected using temperature sampling. Specify a lower value for less random responses and a higher value for more random responses. |
|
|
Top-P changes how the model selects tokens for output. Tokens are selected
from the most (see top-K) to least probable until the sum of their probabilities
equals the top-P value. For example, if tokens A, B, and C have a probability of
0.3, 0.2, and 0.1 and the top-P value is 0.5 , then the model will
select either A or B as the next token by using temperature and excludes C as a
candidate.
Specify a lower value for less random responses and a higher value for more random responses. |
|
|
Specifies a list of strings that tells the model to stop generating text if one
of the strings is encountered in the response. If a string appears multiple
times in the response, then the response truncates where it's first encountered.
The strings are case-sensitive.
For example, if the following is the returned response when stopSequences isn't specified:
public
static string reverse(string myString)
Then the returned response with stopSequences set to ["Str",
"reverse"] is:
public static string
|
|
|
Grounding lets you reference specific data when using language models. When you ground a model, the model can reference internal, confidential, and otherwise specific data from your repository and include the data in the response. Only data stores from Vertex AI Search are supported. |
Path should follow format: |
|
The number of response variations to return. For each request, you're charged for the
output tokens of all candidates, but are only charged once for the input tokens.
Specifying multiple candidates is a Preview feature that works with
|
|
|
Returns the log probabilities of the top candidate tokens at each generation
step. The model's chosen tokens and log probabilities are always returned at
each step, which might not appear in the list of top candidates. Specify the
number of candidates to return by using an integer value in the range of
1 -5 .
|
|
|
Positive values penalize tokens that repeatedly appear in the generated text, decreasing the
probability of repeating content. Acceptable values are -2.0 —2.0 .
|
|
|
Positive values penalize tokens that already appear in the generated text, increasing the
probability of generating more diverse content. Acceptable values are
-2.0 —2.0 .
|
|
|
When seed is fixed to a specific value, the model makes a best effort to provide
the same response for repeated requests. Deterministic output isn't guaranteed.
Also, changing the model or parameter settings, such as the temperature, can
cause variations in the response even when you use the same seed value. By
default, a random seed value is used.
This is a preview feature. |
|
Sample request
REST
To test a text chat by using the Vertex AI API, send a POST request to the publisher model endpoint.
Before using any of the request data, make the following replacements:
- PROJECT_ID: Your project ID.
HTTP method and URL:
POST https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/chat-bison:predict
Request JSON body:
{ "instances": [{ "context": "CONTEXT", "examples": [ { "input": {"content": "EXAMPLE_INPUT"}, "output": {"content": "EXAMPLE_OUTPUT"} }], "messages": [ { "author": "AUTHOR", "content": "CONTENT", }], }], "parameters": { "temperature": TEMPERATURE, "maxOutputTokens": MAX_OUTPUT_TOKENS, "topP": TOP_P, "topK": TOP_K } }
To send your request, choose one of these options:
curl
Save the request body in a file named request.json
,
and execute the following command:
curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/chat-bison:predict"
PowerShell
Save the request body in a file named request.json
,
and execute the following command:
$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }
Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/chat-bison:predict" | Select-Object -Expand Content
You should receive a JSON response similar to the sample response.
Python
To learn how to install or update the Vertex AI SDK for Python, see Install the Vertex AI SDK for Python. For more information, see the Python API reference documentation.
Node.js
Before trying this sample, follow the Node.js setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Node.js API reference documentation.
To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.
Java
Before trying this sample, follow the Java setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Java API reference documentation.
To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.
Response body
{
"predictions": [
{
"candidates": [
{
"author": string,
"content": string
}
],
"citationMetadata": {
"citations": [
{
"startIndex": integer,
"endIndex": integer,
"url": string,
"title": string,
"license": string,
"publicationDate": string
}
]
},
"logprobs": {
"tokenLogProbs": [ float ],
"tokens": [ string ],
"topLogProbs": [ { map<string, float> } ]
},
"safetyAttributes": {
"categories": [ string ],
"blocked": false,
"scores": [ float ],
"errors": [ int ]
}
}
],
"metadata": {
"tokenMetadata": {
"input_token_count": {
"total_tokens": integer,
"total_billable_characters": integer
},
"output_token_count": {
"total_tokens": integer,
"total_billable_characters": integer
}
}
}
}
Response element | Description |
---|---|
content |
Text content of the chat message. |
candidates |
The chat result generated from given message. |
categories |
The display names of Safety Attribute categories associated with the generated content. Order matches the Scores. |
author |
Author tag for the turn. |
scores |
The confidence scores of the each category, higher value means higher confidence. |
blocked |
A flag indicating if the model's input or output was blocked. |
startIndex |
Index in the prediction output where the citation starts (inclusive). Must be >= 0 and < end_index. |
endIndex |
Index in the prediction output where the citation ends (exclusive). Must be > start_index and < len(output). |
url |
URL associated with this citation. If present, this URL links to the webpage of the source of this citation. Possible URLs include news websites, GitHub repos, etc. |
title |
Title associated with this citation. If present, it refers to the title of the source of this citation. Possible titles include news titles, book titles, etc. |
license |
License associated with this recitation. If present, it refers to the license of the source of this citation. Possible licenses include code licenses, e.g., mit license. |
publicationDate |
Publication date associated with this citation. If present, it refers to the date at which the source of this citation was published. Possible formats are YYYY, YYYY-MM, YYYY-MM-DD. |
safetyAttributes |
A collection of categories and their associated confidence scores. 1-1 mapping to candidates . |
input_token_count |
Number of input tokens. This is the total number of tokens across all messages, examples, and context. |
output_token_count |
Number of output tokens. This is the total number of tokens in content across all candidates in the response. |
tokens |
The sampled tokens. |
tokenLogProbs |
The sampled tokens' log probabilities. |
topLogProb |
The most likely candidate tokens and their log probabilities at each step. |
logprobs |
Results of the `logprobs` parameter. 1-1 mapping to `candidates`. |
Sample response
{
"predictions": [
{
"citationMetadata": {
"citations": []
},
"safetyAttributes": {
"scores": [
0.1
],
"categories": [
"Finance"
],
"blocked": false
},
"candidates": [
{
"author": "AUTHOR",
"content": "RESPONSE"
}
]
}
]
}
Stream response from Generative AI models
The parameters are the same for streaming and non-streaming requests to the APIs.
To view sample code requests and responses using the REST API, see Examples using the streaming REST API.
To view sample code requests and responses using the Vertex AI SDK for Python,
see Examples using Vertex AI SDK for Python for streaming.