Gemini - vector

POST https://api.tikway.ai/v1beta/models/google/{model}:embedContent

cURL

Default

curl --request POST \
  --url https://api.tikway.ai/v1beta/models/google/{model}:embedContent \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "content": {
    "parts": [
      {
        "text": "Hello, world!"
      }
    ]
  }
}
'

200

{
  "candidates": [
    {
      "index": 0,
      "content": {
        "role": "model",
        "parts": [
          {
            "text": "Once upon a time, a magic backpack..."
          }
        ]
      },
      "finishReason": "STOP"
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 10,
    "candidatesTokenCount": 12,
    "totalTokenCount": 22
  },
  "modelVersion": "gemini-2.0-flash"
}

Authorizations

Authorization · string · header · required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json

contents · object[] · required

Current conversation content. A single-round request usually has only one Content; multi-round conversations include historical and latest requests in order.

Minimum array length: 1

Hide child attributes

contents. parts · object[] · required

List of content sections.

Minimum array length: 1

Hide child attributes

contents.parts. text · string

Text input.

contents.parts. inline_data · object

Inline Base64 file data.

Hide child attributes

contents.parts.inline_data. mime_type · string

MIME type, such as image/jpeg.

contents.parts.inline_data. data · string

Base64 encoded data.

contents.parts. file_data · object

A reference to the uploaded file in the File API.

Hide child attributes

contents.parts.file_data. mime_type · string

File MIME type.

contents.parts.file_data. file_uri · string

File URI.

contents.parts. function_call · object

Function calls generated by the previous round of models.

Hide child attributes

contents.parts.function_call. name · string

function name.

contents.parts.function_call. args · object

function parameters.

contents.parts. function_response · object

The execution result of the client's function call.

Hide child attributes

contents.parts.function_response. name · string

function name.

contents.parts.function_response. response · object

Function execution return value.

contents.parts. executable_code · object

Executable code.

Hide child attributes

contents.parts.executable_code. language · string

programming language.

contents.parts.executable_code. code · string

code content.

contents.parts. code_execution_result · object

Code execution results.

Hide child attributes

contents.parts.code_execution_result. outcome · string

Execution result status.

contents.parts.code_execution_result. output · string

Execute output.

contents.parts. thought · boolean

Does this part belong to model thinking content?

contents.parts. thought_signature · string

Think content signatures; keep them as is on postback.

contents. role · enum<string>

Content roles. When omitted, it is usually treated as user.

Available options: user, model

systemInstruction · object

System commands set by developers should currently use the plain text part.

Hide child attributes

systemInstruction. parts · object[] · required

Hide child attributes

systemInstruction.parts. text · string

System command text.

tools · object[]

A list of tools that the model can use, such as function calls or code execution.

Hide child attributes

tools. function_declarations · object[]

List of function declarations.

Hide child attributes

tools.function_declarations. name · string

function name.

tools.function_declarations. description · string

Description of function usage.

tools.function_declarations. parameters · object

Function parameters JSON Schema.

tools.function_declarations. response · object

Function responds to JSON Schema.

tools. code_execution · object

Enable code execution tools.

tools. google_search · object

Enable Google search tools.

tools. url_context · object

Enable the URL context tool.

toolConfig · object

Tool call configuration.

Hide child attributes

toolConfig. functionCallingConfig · object

Hide child attributes

toolConfig.functionCallingConfig. mode · enum<string>

Function calling pattern.

Available options: AUTO, ANY, NONE, VALIDATED

toolConfig.functionCallingConfig. allowedFunctionNames · string[]

Function names that allow automatic calling.

safetySettings · object[]

Security filtering rules; the same category can be configured at most once, overriding the default threshold for that category.

Hide child attributes

safetySettings. category · enum<string>

Security Category.

Available options: HARM_CATEGORY_HATE_SPEECH, HARM_CATEGORY_SEXUALLY_EXPLICIT, HARM_CATEGORY_DANGEROUS_CONTENT, HARM_CATEGORY_HARASSMENT, HARM_CATEGORY_CIVIC_INTEGRITY, HARM_CATEGORY_JAILBREAK

safetySettings. threshold · string

Blocking threshold.

safetySettings. method · string

Safety rating method.

generationConfig · object

Generate and output configurations.

Hide child attributes

generationConfig. temperature · number

Sampling temperature.

generationConfig. topP · number

Kernel sampling threshold.

generationConfig. topK · integer

The number of highest probability candidates retained at each step.

generationConfig. candidateCount · integer

Returns the number of candidate replies.

Required range: x >= 1

generationConfig. maxOutputTokens · integer

The maximum number of output tokens.

generationConfig. stopSequences · string[]

Stop the sequence.

generationConfig. responseMimeType · enum<string>

Output MIME type.

Available options: text/plain, application/json

generationConfig. responseSchema · object

The JSON Schema that application/json output should follow.

generationConfig. responseJsonSchema · object

The raw JSON Schema of the JSON output.

generationConfig. responseModalities · enum<string>[]

The requested output modal.

Available options: TEXT, IMAGE, AUDIO

generationConfig. thinkingConfig · object

Hide child attributes

generationConfig.thinkingConfig. thinkingBudget · integer

Think about token budgeting.

generationConfig.thinkingConfig. includeThoughts · boolean

Whether to include a reflection summary in the response.

generationConfig. mediaResolution · string

Media input resolution settings.

cachedContent · string

The cached content name to use as context, in the format cachedContents/{cachedContent}.

serviceTier · string

Request service level.

store · boolean

Whether to log this request; if set, it will override the project-level log configuration.

Response

200 - application/json

Gemini API: models.generateContent normal response. Fully inline and suitable for importing into Apifox.

candidates · object[]

A list of candidate responses generated by the model.

Hide child attributes

candidates. index · integer

Candidate reply index.

candidates. content · object

Candidate reply content.

Hide child attributes

candidates.content. role · enum<string>

Role, fixed as model.

Available options: model

candidates.content. parts · object[]

Reply content part.

Hide child attributes

candidates.content.parts. text · string

Generated text.

candidates.content.parts. inline_data · object

Media data generated inline.

Hide child attributes

candidates.content.parts.inline_data. mime_type · string

candidates.content.parts.inline_data. data · string

Base64 encoded data.

candidates.content.parts. file_data · object

Hide child attributes

candidates.content.parts.file_data. mime_type · string

candidates.content.parts.file_data. file_uri · string

candidates.content.parts. function_call · object

The model request calls the function.

Hide child attributes

candidates.content.parts.function_call. name · string

function name.

candidates.content.parts.function_call. args · object

function parameters.

candidates.content.parts.function_call. id · string

Function call ID.

candidates.content.parts. executable_code · object

Hide child attributes

candidates.content.parts.executable_code. language · string

candidates.content.parts.executable_code. code · string

candidates.content.parts. code_execution_result · object

Hide child attributes

candidates.content.parts.code_execution_result. outcome · string

candidates.content.parts.code_execution_result. output · string

candidates.content.parts. thought · boolean

Whether it is the content of thinking.

candidates.content.parts. thought_signature · string

Think about signing.

candidates. finishReason · string

The ending reason for the candidate reply.

candidates. finishMessage · string

A description of the reason for ending.

candidates. safetyRatings · object[]

The security rating of the candidate reply.

Hide child attributes

candidates.safetyRatings. category · string

candidates.safetyRatings. probability · string

candidates.safetyRatings. blocked · boolean

candidates. citationMetadata · object

Reference metadata.

candidates. tokenCount · integer

Number of candidate reply tokens.

candidates. avgLogprobs · number

Average log probability of candidate responses.

candidates. logprobsResult · object

Term log probability results.

candidates. urlContextMetadata · object

URL context metadata.

promptFeedback · object

Feedback when prompt words are blocked or filtered.

Hide child attributes

promptFeedback. blockReason · string

Blocking reasons.

promptFeedback. blockReasonMessage · string

Description of the reason for blocking.

promptFeedback. safetyRatings · string[]

Prompt word safety rating.

usageMetadata · object

Word usage statistics.

Hide child attributes

usageMetadata. promptTokenCount · integer

Prompt word number.

usageMetadata. candidatesTokenCount · integer

Number of candidate reply tokens.

usageMetadata. totalTokenCount · integer

The total number of tokens.

usageMetadata. cachedContentTokenCount · integer

The number of cached content tokens.

usageMetadata. thoughtsTokenCount · integer

Think about word count.

usageMetadata. toolUsePromptTokenCount · integer

Tool usage hint word token number.

modelVersion · string

The actual model version used.

responseId · string

Response unique ID.