Analysis Agent

Perform data analysis on uploaded datasets

The Analysis Agent performs data analysis on uploaded datasets using either Edison AI or BioAgents Data Analysis backends. It executes code, generates visualizations, and returns results with artifacts.

Function Signature

// src/agents/analysis/index.ts
export async function analysisAgent(input: {
  objective: string;
  datasets: Dataset[];
  type: AnalysisType;
  userId: string;
  conversationStateId: string;
  onPollUpdate?: OnPollUpdate;
}): Promise<AnalysisResult>;

type AnalysisType = "EDISON" | "BIO";

type Dataset = {
  filename: string;
  id: string;
  description: string;
  content?: Buffer;
  path?: string;
};

type AnalysisResult = {
  objective: string;
  output: string;
  jobId?: string;
  reasoning?: string[];
  start?: string;
  end?: string;
  artifacts?: Array<AnalysisArtifact>;
};

Backends

Edison (Default)

Deep analysis via Edison AI agent with automatic file upload.

const result = await analysisAgent({
  objective: "Analyze gene expression patterns and identify key regulators",
  datasets: [
    {
      filename: "expression_data.csv",
      id: "dataset-123",
      description: "RNA-seq gene expression matrix",
      content: buffer
    }
  ],
  type: "EDISON",
  userId: "user-456",
  conversationStateId: "conv-789"
});

Configuration:

EDISON_API_URL=https://your-edison-api
EDISON_API_KEY=your_key
EDISON_TASK_TIMEOUT_MINUTES=30

See Edison deployment.

BIO

BioAgents Data Analysis Agent with state-of-the-art performance.

const result = await analysisAgent({
  objective: "Perform statistical analysis of metabolite concentrations",
  datasets: [
    {
      filename: "metabolites.xlsx",
      id: "dataset-456",
      description: "Metabolomics data from mass spectrometry",
      path: "uploads/metabolites.xlsx"
    }
  ],
  type: "BIO",
  userId: "user-456",
  conversationStateId: "conv-789",
  onPollUpdate: (update) => {
    console.log(update.reasoning); // Real-time trace
  }
});

Configuration:

PRIMARY_ANALYSIS_AGENT=bio
DATA_ANALYSIS_API_URL=https://your-analysis-api
DATA_ANALYSIS_API_KEY=your_key
BIO_ANALYSIS_TASK_TIMEOUT_MINUTES=60

See BioAgents Data Analysis.

Benchmark Performance

BioAgents Analysis Agent achieves state-of-the-art performance on BixBench:

Evaluation Mode	Score
Open-Answer	48.78%
Multiple-Choice (with refusal)	55.12%
Multiple-Choice (without refusal)	64.39%

Outperforms Kepler, GPT-5, and others across all modes. See research paper.

Output Format

The agent returns:

• output: Synthesized analysis results with inline citations
• artifacts: Generated files (plots, tables, code)
• jobId: External job ID for tracking
• reasoning: Real-time reasoning trace (BIO backend only)

Example Response

{
  "objective": "Analyze gene expression",
  "output": "Analysis identified 342 differentially expressed genes (FDR < 0.05). Key regulators include TP53 and MYC with log2 fold changes of -2.3 and 1.8 respectively. See plot_volcano.png for visualization.",
  "jobId": "bio-task-123",
  "artifacts": [
    {
      "id": "artifact-1",
      "filename": "plot_volcano.png",
      "type": "image/png",
      "path": "artifacts/conv-789/artifact-1/plot_volcano.png"
    },
    {
      "id": "artifact-2",
      "filename": "deg_table.csv",
      "type": "text/csv",
      "path": "artifacts/conv-789/artifact-2/deg_table.csv"
    }
  ]
}

Usage Example

// src/routes/deep-research/start.ts
const analysisTasks = plan.filter(t => t.type === "ANALYSIS");

for (const task of analysisTasks) {
  const result = await analysisAgent({
    objective: task.objective,
    datasets: task.datasets,
    type: analysisType,
    userId,
    conversationStateId: conversationState.id,
    onPollUpdate: (update) => {
      // Send real-time updates via WebSocket
      notifyJobUpdate(messageId, update);
    }
  });
  
  task.output = result.output;
  task.jobId = result.jobId;
  task.artifacts = result.artifacts;
}

Configuration

PRIMARY_ANALYSIS_AGENT · string · default: "edison"

Primary agent: edison or bio

EDISON_API_URL · string

Edison API base URL (when PRIMARY_ANALYSIS_AGENT=edison or empty)

EDISON_API_KEY · string

Edison API key

EDISON_TASK_TIMEOUT_MINUTES · number · default: "30"

Edison task timeout in minutes

DATA_ANALYSIS_API_URL · string

BioAgents Analysis API base URL (when PRIMARY_ANALYSIS_AGENT=bio)

DATA_ANALYSIS_API_KEY · string

BioAgents Analysis API key

BIO_ANALYSIS_TASK_TIMEOUT_MINUTES · number · default: "60"

BioAgents Analysis task timeout in minutes

Error Handling

Task timeout

If analysis exceeds timeout, the agent returns an error:

{
  "output": "Error: Analysis task timed out after 60 minutes",
  "jobId": "bio-task-123"
}

Increase BIO_ANALYSIS_TASK_TIMEOUT_MINUTES or EDISON_TASK_TIMEOUT_MINUTES.

Dataset not found

If dataset path is invalid:

{
  "output": "Error: Dataset 'data.csv' not found at path 'uploads/data.csv'"
}

Verify dataset was uploaded successfully and path is correct.

API connection error

If external API is unreachable:

{
  "output": "Error: Failed to connect to analysis API: ECONNREFUSED"
}

Check API URL and network connectivity.

Related

File Upload

Upload datasets for analysis

Planning Agent

Generates analysis tasks

Discovery Agent

Extracts discoveries from results