mirror of
https://github.com/Kilo-Org/kilocode.git
synced 2026-08-30 17:14:40 +08:00
ef6b152ff8
* feat(worktree): add managed workspace cloning (#30117) * test(tui): skip crashing keymap textarea renderer * fix(core): allow skipping migration execution * fix(opencode): remove automatic full session diffs (#30127) * chore: generate * refactor(worktree): move project out of repository * zen: deepseek flash * fix(tui): remount session view on session switch (#30129) Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com> * go: minimax m3 * refactor(opencode): inline local provider helpers (#30169) * refactor(opencode): simplify provider setup flow (#30173) * fix(app): show project sessions before path sync resolves (#30167) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com> * fix(core): preserve session metadata migration identity (#30176) * refactor(session): align namespace imports and inline trivial helpers (#30180) * opencode(run): add queued prompt management (#30103) Direct run mode previously made submitted follow-up prompts irrevocable while a response was still running. Let users edit or remove queued prompts before dispatch without interrupting the active turn. * chore: generate * fix(acp): honor session/cancel by aborting the running turn (#30145) Co-authored-by: Shoubhit Dash <shoubhit2005@gmail.com> * fix(tui): prevent prompt corruption when pasting near wide characters (#29710) Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com> Co-authored-by: Simon Klee <hello@simonklee.dk> * fix(opencode): avoid nullable webfetch format schema (#30215) * chore: generate * fix(core): contain lsp warmup defects (#30226) * add run --replay mode (#30239) * chore: generate * chore: update nix node_modules hashes * fix(stats): restore leaderboard spacing * fix(stats): center top models dot grid * fix(stats): stabilize top models hover * fix(stats): align big-pickle provider resolution (#30274) * feat(app): v2 desktop UI improvements (#29689) Co-authored-by: Brendan Allan <git@brendonovich.dev> Co-authored-by: Brendan Allan <14191578+Brendonovich@users.noreply.github.com> * chore: generate * fix(tui): clarify inline subagent rows (#30051) * fix(tui): handle events across workspaces (#30281) * feat(core): update Copilot for token-based billing (#30181) * fix(tui): keep background marker with subagent label (#30271) * fix(tui): keep retry attempt before message (#30275) * chore: generate * fix(opencode): enforce storage path invariants (#29666) * chore: generate * feat(core): add location-based permission service (#30287) * chore: generate * fix(tui): preserve live parts during session hydration (#30300) * fix(app): restore deferred MCP status updates (#30220) * fix: export v2 stylesheets and declare core node types (#30312) * chore: update nix node_modules hashes * fix(app): avoid suspending on pending child path (#30314) * fix(opencode): remove sunsetted gpt-5.2 and gpt-5.3-codex from allowed models for codex subscriptions (#30316) * chore: generate * feat(core): expose session location * chore: generate * fix(opencode): preserve websocket api errors (#30321) * refactor(core): simplify session pagination * feat(core): add location filesystem contract * feat(core): add dummy location filesystem layer * chore: generate * feat(opencode): add filesystem read and list routes * chore: generate * infra: stats * sync * feat(app): inset new layout session panels (#30342) * fix(app): tab title truncation and close button positioning (#30349) * tui: show model context in run footer (#30380) * tui: revert OpenTUI upgrade to 0.2.16 (#30383) * chore: update nix node_modules hashes * feat(core): add managed repository cache (#30408) * chore: generate * chore: generate * sync * feat(stats): add cache ratio section * feat(core): add flagged project references (#30414) * chore: generate * feat(core): support named migrations (#30418) * fix(stats): clean retired provider rows during sync (#30420) * fix(stats): mention opencode go in top models copy * feat(core): expose project reference filesystem access (#30423) * chore: generate * sync * fix(tui): scope diff viewer to session directory (#30426) * test: widen provider header timeout margin (#30427) * fix(plugin): restore private git install fallback (#30430) * fix(stats): remove leaderboard nav link * chore(opencode): remove scout agent (#30435) * chore: generate * feat(stats): improve cache ratio chart * chore: generate * fix(effect-drizzle-sqlite): preserve transaction begin errors (#30448) * chore: bump effect beta to 74 (#30449) * Revert "tui: revert OpenTUI upgrade to 0.2.16 (#30383)" (#30452) * chore: update nix node_modules hashes * refactor(opencode): improve startup time by 38% (#30453) Co-authored-by: starptech <starptech@starptechs-MBP.fritz.box> * chore: generate * fix(opencode): patch empty Gemini replay messages (#30463) * chore: generate * refactor(core): consolidate filesystem services (#30447) * chore: generate * run: enable interactive replay by default (#30465) * chore: update nix node_modules hashes * refactor(opencode): remove JSON storage migration (#30461) * chore: generate * chore: update nix node_modules hashes * fix(tui): stop idle background task spinner (#30484) * refactor(core): move v1 schemas into core (#30473) * chore: generate * fix: task id passed to background job for continuation (#30485) * chore: generate * feat(core): project copying and tracking directories (#30139) * chore: generate * fix(opencode): preserve signed thinking during anthropic reorder (#30182) * Revert "fix(opencode): preserve signed thinking during anthropic reorder" (#30502) * fix: rm tool reorder logic from old bug (#30483) * chore: generate * feat(app): polish home projects list UI (#30436) * feat(app): polish select-v2 component (#30446) Co-authored-by: Brendan Allan <git@brendonovich.dev> * fix(github): enforce existing git author identity (#30507) * feat(app): new update button (#30460) Co-authored-by: Brendan Allan <git@brendonovich.dev> * fix(opencode): fallback to sh for curl upgrade (#30499) Co-authored-by: Shoubhit Dash <shoubhit2005@gmail.com> * fix(ui): render whole-file patches as complete diffs (#30516) * chore: generate * feat(app): add servers tab to settings dialog (#29675) * refactor(core): consolidate pty service (#30537) * chore: generate * tui: truncate sidebar file paths (#30531) * chore: update nix node_modules hashes * feat(stats): add geo breakdown (#30456) * chore: generate * chore: update nix node_modules hashes * fix(acp): classify apply_patch as edit (#30564) * fix(acp): classify task as think (#30565) * fix(acp): include external directory permission context (#30567) * fix(acp): clean read tool display content (#30569) * fix(tui): route question responses by session directory (#30578) * fix(stats): serve stats og image from banner * docs(go): add Qwen3.7 Plus model (#30594) * fix(openai): preserve websocket idle state (#30586) * refactor(core): remove ai sdk option fields (#30581) * chore: generate * test(core): cover v1 provider option lowering (#30599) * chore: generate * refactor(core): nest model api id (#30603) * fix(core): expose azure openai xhigh efforts (#30620) * feat(core): add skill registry and file agent loading (#30617) * chore: generate * chore: update nix node_modules hashes * fix(stats): count all go usage * chore: remove zed extension and automation (#30628) * fix(opencode): preserve variant for delegated tasks (#30630) * zen: update nvidia tos * fix(opencode): route SAP AI Core reasoning variants through modelParams (#30482) * chore: generate * fix(app): hide unavailable titlebar update (#30642) * feat(app): v2 thinking level selector (#30646) * fix(app,ui): session review reactivity and VCS query cache (#30660) * feat(core): add embedded v2 session runtime and tool foundation (#30632) * chore: generate * chore: update nix node_modules hashes * docs: correct compaction prune default (#30670) * fix(opencode): avoid shell cancel race (#30641) * feat: bump bedrock and add proper mantle support for openai models through aws bedrock (#30464) * test: wait for shell truncation readiness (#30679) * chore: update nix node_modules hashes * refactor(opencode): clean up task tool prompts (#30687) * feat(core): add command registry (#30624) * chore: generate * fix(acp): replay loaded session transcript (#30645) Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com> Co-authored-by: Shoubhit Dash <shoubhit2005@gmail.com> * fix(core): reset pre-launch session projections (#30728) * feat(tui): improve experimental session switcher (#30738) * fix(opencode): respect disabled auto compaction on overflow (#30749) * zen: nemotron 3 ultra * fix(enterprise): install hono standard validator peer (#30740) Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com> * fix build * chore: update nix node_modules hashes * make scripts executable * fix(tui): show toast when variant_list keybind used with no variants (#30724) * fix(opencode): `ACP.loadSession` should replay all messages (#30761) Co-authored-by: Shoubhit Dash <shoubhit2005@gmail.com> * fix(opencode): attribute task child agent on creation (#30786) * fix(tui): add Vue syntax highlighting (#30802) * fix: bump @openrouter/ai-sdk-provider to 2.9.0 (#30800) * feat(core): moving sessions (#30640) * chore: generate * tweak: background agent prompting to avoid polling issues (#30790) * upgrade opentui to 0.3.2 (#30748) * chore: update nix node_modules hashes * feat(desktop): surface local server startup failures (#30822) * ci: publish * refactor(core): make v2 session inputs event sourced (#30785) * chore: generate * fix(llm): normalize OpenAI function tool schemas * chore: generate * feat(stats): refresh stats routes and homepage (#30419) * fix(stats): sort metric charts by top usage * feat(core): add public native API (#30828) * chore: generate * feat(app): color themes (#30824) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com> * chore: generate * sync release versions for v1.16.0 * feat(core): attach global native tools (#30832) * chore: generate * feat(core): add Snowflake Cortex provider (#29901) Co-authored-by: Cortex Code <noreply@snowflake.com> * chore: generate * feat(core): persist v2 session context epochs (#30789) * chore: generate * feat(tui): allow backgrounding synchronous subagents (#30488) * fix(app): improve tab handling (#30669) * chore: generate * fix(tui): prioritize models slash autocomplete (#30848) * fix(tui): route permission replies to session directory (#30851) Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com> * fix(cli): harden daemon lifecycle (#30844) * chore: generate * feat(app): improve desktop multi-server support (#30678) Co-authored-by: Brendan Allan <git@brendonovich.dev> * chore: generate * fix(app): handle tab overflow and scrolling in titlebar (#30886) * fix(app): tab overflow (#30894) * tui: guard path formatting inputs (#30469) Fixes #27726, #25216, #24856, #24294, #17071, #29164, #24837, #16865, #14279, #29895 * opencode/run: refresh themes after terminal reloads (#30917) * chore: generate * fix(tui): fall back to local cwd when editor spawns in attach mode (#30583) * docs: update Go Qwen tiered pricing (#30936) * chore: generate * feat(tui): add diff hunk navigation (#30935) * chore: rm fuzzy search on references (#30931) * fix: use mapError instead of orDie for context snapshot decoding (#30905) Co-authored-by: Shoubhit Dash <shoubhit2005@gmail.com> * fix(core): recover corrupted models cache (#30947) * chore: bun install (#30968) * fix(opencode): resolve Bedrock hang by using node build conditions (#30873) * fix(workflows): retry nix-hashes compute-hash on transient failure (#30743) * fix(stats): scroll model charts to latest on mobile * fix(opencode): prevent destructive edit matches (#30932) * chore: generate * fix(core): respect v2 default agents (#30969) * chore: generate * test(opencode): remove disposal event wait race (#30971) * test(opencode): remove shell timeout output race (#30974) * fix(opencode): gate reasoning summaries by provider (#30973) * feat(core): admit v2 skill guidance (#30843) * fix(workflows): serialize desktop release uploads (#30978) * fix(stats): add mobile chart end spacing * release: v1.16.2 * refactor: kilo compat for v1.16.2 * fix(opencode): address v1.16.2 merge regressions * chore: update kilo-vscode visual regression baselines * fix(opencode): restore Kilo behavior after v1.16.2 merge * fix(opencode): retry Windows migration cleanup * test(opencode): restore clone and macOS watcher coverage * fix(opencode): address second-pass review for #12099 Preserve imported usage and retry partial JSON migrations. Refresh active dependency patches, remove the obsolete GCP patch, and regenerate Kilo HttpApi branding. --------- Co-authored-by: Dax <mail@thdxr.com> Co-authored-by: Dax Raad <d@ironbay.co> Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com> Co-authored-by: Frank <frank@anoma.ly> Co-authored-by: opencode-agent[bot] <219766164+opencode-agent[bot]@users.noreply.github.com> Co-authored-by: Aiden Cline <63023139+rekram1-node@users.noreply.github.com> Co-authored-by: Michael Hart <mhart@cloudflare.com> Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com> Co-authored-by: Simon Klee <hello@simonklee.dk> Co-authored-by: smagnuso <smagnuso@gmail.com> Co-authored-by: Shoubhit Dash <shoubhit2005@gmail.com> Co-authored-by: Orca丶 <93272799+dauphinYan@users.noreply.github.com> Co-authored-by: Adam <2363879+adamdotdevin@users.noreply.github.com> Co-authored-by: Aarav Sareen <96787824+arvsrn@users.noreply.github.com> Co-authored-by: Brendan Allan <git@brendonovich.dev> Co-authored-by: Brendan Allan <14191578+Brendonovich@users.noreply.github.com> Co-authored-by: Kit Langton <kit.langton@gmail.com> Co-authored-by: James Long <longster@gmail.com> Co-authored-by: Dustin Deus <deusdustin@gmail.com> Co-authored-by: starptech <starptech@starptechs-MBP.fritz.box> Co-authored-by: Ulises Jeremias <ulisescf.24@gmail.com> Co-authored-by: Jack <jack@anoma.ly> Co-authored-by: Jérôme Benoit <jerome.benoit@sap.com> Co-authored-by: Ariane Emory <97994360+ariane-emory@users.noreply.github.com> Co-authored-by: LIU Xinyu <contact@lxy.cc> Co-authored-by: Colin McDonnell <colinmcd94@gmail.com> Co-authored-by: Sebastian <hasta84@gmail.com> Co-authored-by: opencode <opencode@sst.dev> Co-authored-by: Kamesh Sampath <kamesh.sampath@hotmail.com> Co-authored-by: Cortex Code <noreply@snowflake.com> Co-authored-by: pcadena-lila <pcadena@lila.ai> Co-authored-by: weiconghe <46336277+weiconghe@users.noreply.github.com> Co-authored-by: alberto <914199+alblez@users.noreply.github.com> Co-authored-by: kilo-maintainer[bot] <kilo-maintainer[bot]@users.noreply.github.com>
388 lines
12 KiB
TypeScript
388 lines
12 KiB
TypeScript
import { describe, expect, test } from "bun:test"
|
|
import { SessionV1 } from "@opencode-ai/core/v1/session"
|
|
import { Exit, Schema } from "effect"
|
|
import { MessageV2 } from "../../src/session/message-v2"
|
|
import { SessionPrompt } from "../../src/session/prompt"
|
|
import { SessionID, MessageID } from "../../src/session/schema"
|
|
|
|
const decodeFormat = Schema.decodeUnknownExit(SessionV1.Format)
|
|
const decodeUser = Schema.decodeUnknownExit(SessionV1.User)
|
|
const decodeAssistant = Schema.decodeUnknownExit(SessionV1.Assistant)
|
|
|
|
describe("structured-output.OutputFormat", () => {
|
|
test("parses text format", () => {
|
|
const result = decodeFormat({ type: "text" })
|
|
expect(Exit.isSuccess(result)).toBe(true)
|
|
if (Exit.isSuccess(result)) {
|
|
expect(result.value.type).toBe("text")
|
|
}
|
|
})
|
|
|
|
test("parses json_schema format with defaults", () => {
|
|
const result = decodeFormat({
|
|
type: "json_schema",
|
|
schema: { type: "object", properties: { name: { type: "string" } } },
|
|
})
|
|
expect(Exit.isSuccess(result)).toBe(true)
|
|
if (Exit.isSuccess(result)) {
|
|
expect(result.value.type).toBe("json_schema")
|
|
if (result.value.type === "json_schema") {
|
|
expect(result.value.retryCount).toBe(2) // default value
|
|
}
|
|
}
|
|
})
|
|
|
|
test("parses json_schema format with custom retryCount", () => {
|
|
const result = decodeFormat({
|
|
type: "json_schema",
|
|
schema: { type: "object" },
|
|
retryCount: 5,
|
|
})
|
|
expect(Exit.isSuccess(result)).toBe(true)
|
|
if (Exit.isSuccess(result) && result.value.type === "json_schema") {
|
|
expect(result.value.retryCount).toBe(5)
|
|
}
|
|
})
|
|
|
|
test("rejects invalid type", () => {
|
|
const result = decodeFormat({ type: "invalid" })
|
|
expect(Exit.isFailure(result)).toBe(true)
|
|
})
|
|
|
|
test("rejects json_schema without schema", () => {
|
|
const result = decodeFormat({ type: "json_schema" })
|
|
expect(Exit.isFailure(result)).toBe(true)
|
|
})
|
|
|
|
test("rejects negative retryCount", () => {
|
|
const result = decodeFormat({
|
|
type: "json_schema",
|
|
schema: { type: "object" },
|
|
retryCount: -1,
|
|
})
|
|
expect(Exit.isFailure(result)).toBe(true)
|
|
})
|
|
})
|
|
|
|
describe("structured-output.StructuredOutputError", () => {
|
|
test("creates error with message and retries", () => {
|
|
const error = new SessionV1.StructuredOutputError({
|
|
message: "Failed to validate",
|
|
retries: 3,
|
|
})
|
|
|
|
expect(error.name).toBe("StructuredOutputError")
|
|
expect(error.data.message).toBe("Failed to validate")
|
|
expect(error.data.retries).toBe(3)
|
|
})
|
|
|
|
test("converts to object correctly", () => {
|
|
const error = new SessionV1.StructuredOutputError({
|
|
message: "Test error",
|
|
retries: 2,
|
|
})
|
|
|
|
const obj = error.toObject()
|
|
expect(obj.name).toBe("StructuredOutputError")
|
|
expect(obj.data.message).toBe("Test error")
|
|
expect(obj.data.retries).toBe(2)
|
|
})
|
|
|
|
test("isInstance correctly identifies error", () => {
|
|
const error = new SessionV1.StructuredOutputError({
|
|
message: "Test",
|
|
retries: 1,
|
|
})
|
|
|
|
expect(SessionV1.StructuredOutputError.isInstance(error)).toBe(true)
|
|
expect(SessionV1.StructuredOutputError.isInstance({ name: "other" })).toBe(false)
|
|
})
|
|
})
|
|
|
|
describe("structured-output.UserMessage", () => {
|
|
test("user message accepts outputFormat", () => {
|
|
const result = decodeUser({
|
|
id: MessageID.ascending(),
|
|
sessionID: SessionID.descending(),
|
|
role: "user",
|
|
time: { created: Date.now() },
|
|
agent: "default",
|
|
model: { providerID: "anthropic", modelID: "claude-3" },
|
|
outputFormat: {
|
|
type: "json_schema",
|
|
schema: { type: "object" },
|
|
},
|
|
})
|
|
expect(Exit.isSuccess(result)).toBe(true)
|
|
})
|
|
|
|
test("user message works without outputFormat (optional)", () => {
|
|
const result = decodeUser({
|
|
id: MessageID.ascending(),
|
|
sessionID: SessionID.descending(),
|
|
role: "user",
|
|
time: { created: Date.now() },
|
|
agent: "default",
|
|
model: { providerID: "anthropic", modelID: "claude-3" },
|
|
})
|
|
expect(Exit.isSuccess(result)).toBe(true)
|
|
})
|
|
})
|
|
|
|
describe("structured-output.AssistantMessage", () => {
|
|
const baseAssistantMessage = {
|
|
id: MessageID.ascending(),
|
|
sessionID: SessionID.descending(),
|
|
role: "assistant" as const,
|
|
parentID: MessageID.ascending(),
|
|
modelID: "claude-3",
|
|
providerID: "anthropic",
|
|
mode: "default",
|
|
agent: "default",
|
|
path: { cwd: "/test", root: "/test" },
|
|
cost: 0.001,
|
|
tokens: { input: 100, output: 50, reasoning: 0, cache: { read: 0, write: 0 } },
|
|
time: { created: Date.now() },
|
|
}
|
|
|
|
test("assistant message accepts structured", () => {
|
|
const result = decodeAssistant({
|
|
...baseAssistantMessage,
|
|
structured: { company: "Anthropic", founded: 2021 },
|
|
})
|
|
expect(Exit.isSuccess(result)).toBe(true)
|
|
if (Exit.isSuccess(result)) {
|
|
expect(result.value.structured).toEqual({ company: "Anthropic", founded: 2021 })
|
|
}
|
|
})
|
|
|
|
test("assistant message works without structured_output (optional)", () => {
|
|
const result = decodeAssistant(baseAssistantMessage)
|
|
expect(Exit.isSuccess(result)).toBe(true)
|
|
})
|
|
})
|
|
|
|
describe("structured-output.createStructuredOutputTool", () => {
|
|
test("creates tool with description", () => {
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema: { type: "object" },
|
|
onSuccess: () => {},
|
|
})
|
|
|
|
expect(tool.description).toContain("structured format")
|
|
})
|
|
|
|
test("creates tool with schema as inputSchema", () => {
|
|
const schema = {
|
|
type: "object",
|
|
properties: {
|
|
company: { type: "string" },
|
|
founded: { type: "number" },
|
|
},
|
|
required: ["company"],
|
|
}
|
|
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema,
|
|
onSuccess: () => {},
|
|
})
|
|
|
|
// AI SDK wraps schema in { jsonSchema: {...} }
|
|
expect(tool.inputSchema).toBeDefined()
|
|
const inputSchema = tool.inputSchema as any
|
|
expect(inputSchema.jsonSchema?.properties?.company).toBeDefined()
|
|
expect(inputSchema.jsonSchema?.properties?.founded).toBeDefined()
|
|
})
|
|
|
|
test("strips $schema property from inputSchema", () => {
|
|
const schema = {
|
|
$schema: "http://json-schema.org/draft-07/schema#",
|
|
type: "object",
|
|
properties: { name: { type: "string" } },
|
|
}
|
|
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema,
|
|
onSuccess: () => {},
|
|
})
|
|
|
|
// AI SDK wraps schema in { jsonSchema: {...} }
|
|
const inputSchema = tool.inputSchema as any
|
|
expect(inputSchema.jsonSchema?.$schema).toBeUndefined()
|
|
})
|
|
|
|
test("execute calls onSuccess with valid args", async () => {
|
|
let capturedOutput: unknown
|
|
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema: { type: "object", properties: { name: { type: "string" } } },
|
|
onSuccess: (output) => {
|
|
capturedOutput = output
|
|
},
|
|
})
|
|
|
|
expect(tool.execute).toBeDefined()
|
|
const testArgs = { name: "Test Company" }
|
|
const result = await tool.execute!(testArgs, {
|
|
toolCallId: "test-call-id",
|
|
messages: [],
|
|
abortSignal: undefined as any,
|
|
})
|
|
|
|
expect(capturedOutput).toEqual(testArgs)
|
|
expect(result.output).toBe("Structured output captured successfully.")
|
|
expect(result.metadata.valid).toBe(true)
|
|
})
|
|
|
|
test("AI SDK validates schema before execute - missing required field", async () => {
|
|
// Note: The AI SDK validates the input against the schema BEFORE calling execute()
|
|
// So invalid inputs never reach the tool's execute function
|
|
// This test documents the expected schema behavior
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema: {
|
|
type: "object",
|
|
properties: {
|
|
name: { type: "string" },
|
|
age: { type: "number" },
|
|
},
|
|
required: ["name", "age"],
|
|
},
|
|
onSuccess: () => {},
|
|
})
|
|
|
|
// The schema requires both 'name' and 'age'
|
|
expect(tool.inputSchema).toBeDefined()
|
|
const inputSchema = tool.inputSchema as any
|
|
expect(inputSchema.jsonSchema?.required).toContain("name")
|
|
expect(inputSchema.jsonSchema?.required).toContain("age")
|
|
})
|
|
|
|
test("AI SDK validates schema types before execute - wrong type", async () => {
|
|
// Note: The AI SDK validates the input against the schema BEFORE calling execute()
|
|
// So invalid inputs never reach the tool's execute function
|
|
// This test documents the expected schema behavior
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema: {
|
|
type: "object",
|
|
properties: {
|
|
count: { type: "number" },
|
|
},
|
|
required: ["count"],
|
|
},
|
|
onSuccess: () => {},
|
|
})
|
|
|
|
// The schema defines 'count' as a number
|
|
expect(tool.inputSchema).toBeDefined()
|
|
const inputSchema = tool.inputSchema as any
|
|
expect(inputSchema.jsonSchema?.properties?.count?.type).toBe("number")
|
|
})
|
|
|
|
test("execute handles nested objects", async () => {
|
|
let capturedOutput: unknown
|
|
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema: {
|
|
type: "object",
|
|
properties: {
|
|
user: {
|
|
type: "object",
|
|
properties: {
|
|
name: { type: "string" },
|
|
email: { type: "string" },
|
|
},
|
|
required: ["name"],
|
|
},
|
|
},
|
|
required: ["user"],
|
|
},
|
|
onSuccess: (output) => {
|
|
capturedOutput = output
|
|
},
|
|
})
|
|
|
|
// Valid nested object - AI SDK validates before calling execute()
|
|
const validResult = await tool.execute!(
|
|
{ user: { name: "John", email: "john@test.com" } },
|
|
{
|
|
toolCallId: "test-call-id",
|
|
messages: [],
|
|
abortSignal: undefined as any,
|
|
},
|
|
)
|
|
|
|
expect(capturedOutput).toEqual({ user: { name: "John", email: "john@test.com" } })
|
|
expect(validResult.metadata.valid).toBe(true)
|
|
|
|
// Verify schema has correct nested structure
|
|
const inputSchema = tool.inputSchema as any
|
|
expect(inputSchema.jsonSchema?.properties?.user?.type).toBe("object")
|
|
expect(inputSchema.jsonSchema?.properties?.user?.properties?.name?.type).toBe("string")
|
|
expect(inputSchema.jsonSchema?.properties?.user?.required).toContain("name")
|
|
})
|
|
|
|
test("execute handles arrays", async () => {
|
|
let capturedOutput: unknown
|
|
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema: {
|
|
type: "object",
|
|
properties: {
|
|
tags: {
|
|
type: "array",
|
|
items: { type: "string" },
|
|
},
|
|
},
|
|
required: ["tags"],
|
|
},
|
|
onSuccess: (output) => {
|
|
capturedOutput = output
|
|
},
|
|
})
|
|
|
|
// Valid array - AI SDK validates before calling execute()
|
|
const validResult = await tool.execute!(
|
|
{ tags: ["a", "b", "c"] },
|
|
{
|
|
toolCallId: "test-call-id",
|
|
messages: [],
|
|
abortSignal: undefined as any,
|
|
},
|
|
)
|
|
|
|
expect(capturedOutput).toEqual({ tags: ["a", "b", "c"] })
|
|
expect(validResult.metadata.valid).toBe(true)
|
|
|
|
// Verify schema has correct array structure
|
|
const inputSchema = tool.inputSchema as any
|
|
expect(inputSchema.jsonSchema?.properties?.tags?.type).toBe("array")
|
|
expect(inputSchema.jsonSchema?.properties?.tags?.items?.type).toBe("string")
|
|
})
|
|
|
|
test("toModelOutput returns text value", async () => {
|
|
const tool = SessionPrompt.createStructuredOutputTool({
|
|
schema: { type: "object" },
|
|
onSuccess: () => {},
|
|
})
|
|
|
|
expect(tool.toModelOutput).toBeDefined()
|
|
const modelOutput = await Promise.resolve(
|
|
tool.toModelOutput!({
|
|
toolCallId: "test-call-id",
|
|
input: {},
|
|
output: {
|
|
output: "Test output",
|
|
},
|
|
}),
|
|
)
|
|
|
|
expect(modelOutput.type).toBe("text")
|
|
if (modelOutput.type !== "text") throw new Error("expected text model output")
|
|
expect(modelOutput.value).toBe("Test output")
|
|
})
|
|
|
|
// Note: Retry behavior is handled by the AI SDK and the prompt loop, not the tool itself
|
|
// The tool simply calls onSuccess when execute() is called with valid args
|
|
// See prompt.ts loop() for actual retry logic
|
|
})
|