Repository intelligence

TypeScript SDK reference for repository intelligence — snapshot a repo, triage an issue into ranked evidence, build a decision plan, query the symbol graph, simulate, and apply a fix.

Repository intelligence turns a code repository into governed, queryable evidence. With AlgentaClient you snapshot the repo, triage an issue into ranked evidence, build a decision plan from that evidence, walk the symbol graph, simulate a candidate plan, and apply the result as a patch, a local branch, or a pull request. Method names are camelCase; request bodies stay snake_case, mirroring the HTTP contract. See the repository intelligence guide for the end-to-end story.

import { AlgentaClient } from "algenta-sdk";

const client = new AlgentaClient({
  apiKey: process.env.ALGENTA_API_KEY ?? process.env.DE_API_KEY,
  baseUrl: process.env.ALGENTA_BASE_URL ?? "https://api.algenta.ai",
});

Check support, then snapshot

Start with getRepositoryIntelligenceCapabilities to see which languages the engine supports, then take a snapshot with createRepositorySnapshot. A snapshot returns a snapshot_id plus content hashes and per-artifact references you reuse for every later call.

TypeScript

const capabilities = await client.getRepositoryIntelligenceCapabilities();
console.log(
  capabilities.supported_languages.length,
  capabilities.support_progress.progress_label,
);

const snapshot = await client.createRepositorySnapshot("$REPOSITORY_ID", {
  ref: "main",
  include_patterns: ["src/**"],
  max_files: 5_000,
});
console.log(snapshot.snapshot_id, snapshot.file_count, snapshot.status);

// Re-read a snapshot later by id.
const again = await client.getRepositorySnapshot("$REPOSITORY_ID", snapshot.snapshot_id);
console.log(again.resolved_revision, again.content_hash);

cURL

curl -sS "$ALGENTA_BASE_URL/v1/repositories/$REPOSITORY_ID/snapshots" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "ref": "main", "include_patterns": ["src/**"], "max_files": 5000 }'

Triage, plan, and inspect the graph

triageRepository reduces the whole repo to ranked evidence for a set of signals (issue_text, failing_tests, changed_files, diagnostics, workspace_context) and returns a workspace_evidence_bundle_ref. Feed that bundle into createRepositoryDecisionPlan to get a decision_plan_id, and use queryRepositoryGraph to understand blast radius around a file or symbol.

// 1. Triage the issue into ranked evidence.
const triage = await client.triageRepository("$REPOSITORY_ID", {
  snapshot_id: snapshot.snapshot_id,
  signals: {
    issue_text: "Null pointer when the cart is empty at checkout",
    failing_tests: ["test_checkout_empty_cart"],
  },
  max_evidence_items: 20,
});
console.log(triage.suspect_files, triage.reduction_ratio, triage.evidence_items.length);

// 2. Build a decision plan from the evidence bundle.
const plan = await client.createRepositoryDecisionPlan("$REPOSITORY_ID", {
  snapshot_id: snapshot.snapshot_id,
  workspace_evidence_bundle_ref: triage.workspace_evidence_bundle_ref,
});
console.log(plan.decision_plan_id);

// 3. Walk the symbol/file graph around a symbol.
const graph = await client.queryRepositoryGraph("$REPOSITORY_ID", {
  snapshot_id: snapshot.snapshot_id,
  symbol_name: "checkout",
  direction: "both", // "inbound" | "outbound" | "both"
  max_depth: 2,
});
console.log(graph.impacted_files, graph.top_change_risk_files.length);

Simulate, then apply

simulateRepository scores a candidate decision_plan_id and returns a DecisionEnvelope. The envelope exposes the simulation_id in validated_inputs — pass that to applyRepository, which writes the result according to mode.

const decision = await client.simulateRepository("$REPOSITORY_ID", {
  snapshot_id: snapshot.snapshot_id,
  decision_plan_id: plan.decision_plan_id,
  runs: 256,
});
const simulationId = decision.validated_inputs?.simulation_id as string;
console.log(decision.recommended_action, decision.confidence, simulationId);

const applied = await client.applyRepository("$REPOSITORY_ID", {
  snapshot_id: snapshot.snapshot_id,
  decision_plan_id: plan.decision_plan_id,
  simulation_id: simulationId, // from the simulate step's validated_inputs
  mode: "remote_pr", // "patch_only" | "local_branch" | "remote_pr"
  write_permission: true,
  pull_request_title: "Fix empty-cart checkout crash",
});
console.log(applied.applied, applied.pull_request_url, applied.commit_sha);

applyRepository with mode: "remote_pr" or "local_branch" writes to your repository. Both require write_permission: true and a simulation_id from a prior simulateRepository call. Use mode: "patch_only" to get the unified diff in patch without writing anything.

Method reference

Method	HTTP	Notes
getRepositoryIntelligenceCapabilities()	GET /v1/repositories/capabilities	supported_languages and support_progress.
createRepositorySnapshot(repositoryId, request)	POST /v1/repositories/{id}/snapshots	RepositorySnapshotCreateRequest (ref, include_patterns, exclude_patterns, max_files, max_file_size_bytes).
getRepositorySnapshot(repositoryId, snapshotId)	GET /v1/repositories/{id}/snapshots/{snapshot_id}	Re-read a snapshot by id.
triageRepository(repositoryId, request)	POST /v1/repositories/{id}/triage	RepositoryTriageRequest (snapshot_id, signals, max_evidence_items, max_snippet_lines, token_budget); returns ranked evidence_items.
createRepositoryDecisionPlan(repositoryId, request)	POST /v1/repositories/{id}/decision-plans	RepositoryDecisionPlanCreateRequest (workspace_evidence_bundle_ref required, snapshot_id, model).
queryRepositoryGraph(repositoryId, request)	POST /v1/repositories/{id}/graph-query	RepositoryGraphQueryRequest (snapshot_id, file_path, symbol_name, direction, max_depth, max_nodes).
simulateRepository(repositoryId, request)	POST /v1/repositories/{id}/simulate	RepositorySimulationRequest (decision_plan_id required, snapshot_id, runs, seed); returns a DecisionEnvelope.
applyRepository(repositoryId, request)	POST /v1/repositories/{id}/apply	RepositoryApplyRequest (decision_plan_id, simulation_id, mode required; write_permission, branch_name, commit_message, base_branch, pull_request_title, pull_request_body).

Related pages

Triage a repositoryAgent runsTypeScript SDK