@voltro/ai 0.86.0 → 0.88.0

package/dist/realtimeEvent.js

@@ -1,5 +1,5 @@
 import { Schema as e } from "effect";
-//#region src/realtimeEvent.ts
+// src/realtimeEvent.ts
 var t = e.String.pipe(e.check(e.isPattern(/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/))), n = e.Number.pipe(e.check(e.isFinite()), e.check(e.isGreaterThanOrEqualTo(0))), r = e.Union([
 	e.Struct({
 		type: e.Literal("input-audio-append"),
@@ -93,5 +93,4 @@ var t = e.String.pipe(e.check(e.isPattern(/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/
 		transcript: e.optional(e.String)
 	})
 ]);
-//#endregion
 export { r as RealtimeClientEvent, o as RealtimeServerEvent };

package/dist/realtimeSession.js

@@ -4,7 +4,7 @@ import { createGateway as r } from "@ai-sdk/gateway";
 import { Effect as i, Queue as a, Result as o, Schema as s } from "effect";
 import { defineWebSocket as c } from "@voltro/protocol";
 import { WebSocket as l } from "ws";
-//#region src/realtimeSession.ts
+// src/realtimeSession.ts
 var u = (e) => {
 	if (e.name !== "gateway" && e.name !== "openai") throw TypeError(`Provider '${e.name}' does not support realtime sessions. Use 'gateway' or 'openai'.`);
 	let t = {
@@ -150,5 +150,4 @@ var u = (e) => {
 	let i = JSON.stringify(t);
 	return i === void 0 || Buffer.byteLength(i) > n.maxMessageBytes || e.bufferedAmount + Buffer.byteLength(i) > n.maxPendingBytes ? Promise.reject(new p(4409, "Realtime buffer limit exceeded")) : new Promise((t, n) => e.send(i, (e) => e ? n(new p(4502, "Realtime provider failed")) : t()));
 };
-//#endregion
 export { h as defineRealtimeSession, u as resolveRealtimeModel };

package/dist/rerank.js

@@ -1,7 +1,7 @@
 import { aiGenerationError as e } from "./api.js";
 import { rerank as t } from "ai";
 import { Effect as n } from "effect";
-//#region src/rerank.ts
+// src/rerank.ts
 var r = (t) => n.tryPromise({
 	try: async () => {
 		let e = t.getText ?? ((e) => String(e)), n = t.provider ?? "mock", r = [...typeof n == "string" ? n === "mock" ? i(t.query, t.documents, e) : await o(n, t.query, t.documents, e, t.model) : await s(n, t.query, t.documents, e, t.model)].sort((e, t) => t.score - e.score);
@@ -57,5 +57,4 @@ var r = (t) => n.tryPromise({
 		score: e.score
 	}));
 };
-//#endregion
 export { r as rerank };

package/dist/resumableStream.js

@@ -2,7 +2,7 @@ import { streamText as e } from "./streamText.js";
 import { awaitStreamChange as t, hasStreamReaders as n, watchStream as r, withStreamWakeups as i } from "./streamWakeups.js";
 import { Cause as a, Clock as o, Duration as s, Effect as c, Exit as l, Option as u, Stream as d } from "effect";
 import { and as f, boolean as p, eq as m, generateId as h, gt as g, id as _, inSet as v, integer as y, json as b, lt as x, registerRetention as S, retentionTtlMsFromEnv as C, table as w, text as T, timestamp as E } from "@voltro/database";
-//#region src/resumableStream.ts
+// src/resumableStream.ts
 var D = (e) => e._tag === "done" || e._tag === "error", O = (e, n, i = {}) => {
 	let a = i.pollMs ?? 50, l = Math.max(a, i.maxPollMs ?? 1e3);
 	if (!Number.isFinite(a) || a <= 0 || !Number.isFinite(l)) throw RangeError("Stream poll intervals must be finite positive milliseconds");
@@ -227,7 +227,7 @@ var D = (e) => e._tag === "done" || e._tag === "error", O = (e, n, i = {}) => {
 			seq: e.seq,
 			event: e.event,
 			createdAt: r
-		})));
+		})), { returning: "none" });
 	}),
 	read: (t, n, r) => c.promise(async () => (await e.query({
 		table: N,
@@ -317,5 +317,4 @@ var D = (e) => e._tag === "done" || e._tag === "error", O = (e, n, i = {}) => {
 		isDone: (t) => c.promise(() => e.exists(s(t)))
 	});
 };
-//#endregion
 export { L as DEFAULT_STREAM_LOG_RETENTION_HOURS, N as STREAM_EVENTS_TABLE, P as STREAM_STATE_TABLE, O as consumeResumableStream, V as dataStoreResumableStreamStore, H as gcResumableStreams, M as memoryResumableStreamStore, U as redisResumableStreamStore, A as resumableStream, j as resumableStreamText, k as startResumableProducer, F as streamEventsTable, I as streamStateTable };

package/dist/retrieveReranked.js

@@ -1,6 +1,6 @@
 import { rerank as e } from "./rerank.js";
 import { Effect as t } from "effect";
-//#region src/retrieveReranked.ts
+// src/retrieveReranked.ts
 var n = (n) => t.gen(function* () {
 	let t = n.candidates ?? 50, r = n.topN ?? 8, i = yield* n.search(t);
 	return i.length === 0 ? i : (yield* e({
@@ -12,5 +12,4 @@ var n = (n) => t.gen(function* () {
 		...n.model === void 0 ? {} : { model: n.model }
 	})).map((e) => e.document);
 });
-//#endregion
 export { n as retrieveReranked };

package/dist/runAndPersist.js

@@ -3,7 +3,7 @@ import { recordAiUsageBestEffort as t } from "./cost.js";
 import { streamText as n } from "./streamText.js";
 import { AGENT_MESSAGES_TABLE as r, AGENT_THREADS_TABLE as i } from "./agentConstants.js";
 import { Effect as a, Stream as o } from "effect";
-//#region src/runAndPersist.ts
+// src/runAndPersist.ts
 var s = (e) => [...e].sort((e, t) => e.order - t.order || e.stepOrder - t.stepOrder).filter((e) => !e.streaming && e.content.length > 0 && (e.role === "user" || e.role === "assistant")).map((e) => e.role === "user" ? {
 	role: "user",
 	content: e.content
@@ -29,55 +29,58 @@ var s = (e) => [...e].sort((e, t) => e.order - t.order || e.stepOrder - t.stepOr
 		content: u(n)
 	});
 }), u = (e) => e.filter((e) => e.type === "text").map((e) => e.text).join(""), d = (i, s) => a.gen(function* () {
-	let d = yield* c(i, {
+	let d = s.provider ?? (yield* a.try({
+		try: () => e(),
+		catch: (e) => e instanceof Error ? e : Error(String(e))
+	})), f = yield* c(i, {
 		threadId: s.threadId,
 		tenantId: s.tenantId ?? null,
 		order: s.order
-	}), f = "", p = "", m = /* @__PURE__ */ new Map(), h = [], g = [], _ = [], v = 0, y = null, b, x = s.throttleMs ?? 100, S = () => {
+	}), p = "", m = "", h = /* @__PURE__ */ new Map(), g = [], _ = [], v = [], y = 0, b = null, x, S = s.throttleMs ?? 100, C = () => {
 		let e = [];
-		f.length > 0 && e.push({
+		p.length > 0 && e.push({
 			type: "reasoning",
-			text: f
-		}), p.length > 0 && e.push({
-			type: "text",
 			text: p
+		}), m.length > 0 && e.push({
+			type: "text",
+			text: m
 		});
-		for (let t of m.values()) e.push(t);
-		for (let t of h) e.push(t);
+		for (let t of h.values()) e.push(t);
 		for (let t of g) e.push(t);
 		for (let t of _) e.push(t);
+		for (let t of v) e.push(t);
 		return e;
-	}, C = (e) => {
+	}, w = (e) => {
 		let t = Date.now();
-		return !e && t - v < x ? a.void : (v = t, l(i, d, S()));
+		return !e && t - y < S ? a.void : (y = t, l(i, f, C()));
 	};
 	yield* n({
 		prompt: s.prompt,
 		...s.messages === void 0 ? {} : { messages: s.messages },
 		...s.system === void 0 ? {} : { system: s.system },
 		...s.tools === void 0 ? {} : { tools: s.tools },
-		...s.provider === void 0 ? {} : { provider: s.provider },
+		provider: d,
 		...s.maxSteps === void 0 ? {} : { maxSteps: s.maxSteps },
 		...s.toolContext === void 0 ? {} : { toolContext: s.toolContext },
 		...s.providerOptions === void 0 ? {} : { providerOptions: s.providerOptions }
 	}).pipe(o.runForEach((e) => a.gen(function* () {
 		switch (e._tag) {
 			case "token":
-				p += e.text, yield* C(!1);
+				m += e.text, yield* w(!1);
 				break;
 			case "toolCall":
-				m.set(e.id, {
+				h.set(e.id, {
 					type: "tool",
 					toolCallId: e.id,
 					toolName: e.name,
 					state: "input-available",
 					input: e.input,
 					...e.execution === void 0 ? {} : { execution: e.execution }
-				}), yield* C(!0);
+				}), yield* w(!0);
 				break;
 			case "toolResult": {
-				let t = m.get(e.id);
-				m.set(e.id, {
+				let t = h.get(e.id);
+				h.set(e.id, {
 					type: "tool",
 					toolCallId: e.id,
 					toolName: e.name,
@@ -85,64 +88,59 @@ var s = (e) => [...e].sort((e, t) => e.order - t.order || e.stepOrder - t.stepOr
 					input: t?.input,
 					output: e.output,
 					...e.error ? { errorText: e.error } : {}
-				}), yield* C(!0);
+				}), yield* w(!0);
 				break;
 			}
 			case "reasoning":
-				f += e.text, yield* C(!1);
+				p += e.text, yield* w(!1);
 				break;
 			case "source": {
 				let { _tag: t, ...n } = e;
-				h.push({
+				g.push({
 					type: t,
 					...n
-				}), yield* C(!0);
+				}), yield* w(!0);
 				break;
 			}
 			case "file": {
 				let { _tag: t, ...n } = e;
-				g.push({
+				_.push({
 					type: t,
 					...n
-				}), yield* C(!0);
+				}), yield* w(!0);
 				break;
 			}
 			case "stepMetadata": {
 				let { _tag: t, ...n } = e;
-				_.push({
+				v.push({
 					type: t,
 					...n
-				}), yield* C(!0);
+				}), yield* w(!0);
 				break;
 			}
 			case "error":
-				y = e.message;
+				b = e.message;
 				break;
-			case "done": b = e.usage;
+			case "done": x = e.usage;
 		}
 	})));
-	let w = S();
-	if (yield* a.promise(() => i.update(r, d, {
-		parts: w,
-		content: u(w),
+	let T = C();
+	return yield* a.promise(() => i.update(r, f, {
+		parts: T,
+		content: u(T),
 		streaming: !1
-	})), s.recordUsage !== !1 && b !== void 0) {
-		let n = s.provider ?? e();
-		yield* t(i, {
-			tenantId: s.tenantId ?? null,
-			provider: n.name,
-			model: n.model ?? "unknown",
-			operation: "agent.turn",
-			agent: s.agent ?? null,
-			requestId: d,
-			usage: {
-				inputTokens: b.inputTokens,
-				outputTokens: b.outputTokens,
-				...b.totalTokens === void 0 ? {} : { totalTokens: b.totalTokens }
-			}
-		});
-	}
-	return y ? yield* a.fail(Error(y)) : u(w);
+	})), s.recordUsage !== !1 && x !== void 0 && (yield* t(i, {
+		tenantId: s.tenantId ?? null,
+		provider: d.name,
+		model: d.model ?? "unknown",
+		operation: "agent.turn",
+		agent: s.agent ?? null,
+		requestId: f,
+		usage: {
+			inputTokens: x.inputTokens,
+			outputTokens: x.outputTokens,
+			...x.totalTokens === void 0 ? {} : { totalTokens: x.totalTokens }
+		}
+	})), b ? yield* a.fail(Error(b)) : u(T);
 });
-//#endregion
 export { c as appendStreamingMessage, s as historyToModelMessages, l as patchStreamingMessage, d as runAssistant };

package/dist/semanticCache.js

@@ -2,7 +2,7 @@ import { generateObject as e, generateText as t } from "./api.js";
 import { embed as n } from "./embed.js";
 import { Effect as r, Option as i } from "effect";
 import { rowDep as a, tableDep as o } from "@voltro/cache";
-//#region src/semanticCache.ts
+// src/semanticCache.ts
 var s = {
 	inputTokens: 0,
 	outputTokens: 0,
@@ -62,5 +62,4 @@ var s = {
 		reset: () => t.clear()
 	};
 };
-//#endregion
 export { m as recordReads, p as semanticGenerateObject, f as semanticGenerateText };

package/dist/speechProvider.js

@@ -1,4 +1,4 @@
-//#region src/speechProvider.ts
+// src/speechProvider.ts
 var e = Symbol.for("@voltro/ai/speechProviders"), t = globalThis, n = () => t[e] ?? (t[e] = /* @__PURE__ */ new Map()), r = [
 	"mock",
 	"openai",
@@ -12,5 +12,4 @@ var e = Symbol.for("@voltro/ai/speechProviders"), t = globalThis, n = () => t[e]
 }, a = (e) => r.includes(e) && n().has(e), o = (e) => n().get(e), s = () => [...n().keys()].sort(), c = () => {
 	n().clear();
 };
-//#endregion
 export { r as BUILTIN_SPEECH_PROVIDERS, c as clearSpeechProviders, i as registerSpeechProvider, s as registeredSpeechProviders, a as replacesBuiltinSpeechProvider, o as speechProvider };

package/dist/streamRegistry.js

@@ -1,4 +1,4 @@
-//#region src/streamRegistry.ts
+// src/streamRegistry.ts
 var e = /* @__PURE__ */ new Map(), t = (t) => {
 	let n = e.get(t);
 	if (n) return n;
@@ -10,5 +10,4 @@ var e = /* @__PURE__ */ new Map(), t = (t) => {
 }, r = (t) => e.has(t), i = (t) => {
 	e.delete(t);
 }, a = () => e.size;
-//#endregion
 export { a as activeStreamCount, n as cancelStream, r as isStreamActive, t as registerStream, i as unregisterStream };

package/dist/streamText.js

@@ -4,7 +4,7 @@ import { aiCallAttributes as r, observeStream as i } from "./observe.js";
 import { registerStream as a, unregisterStream as o } from "./streamRegistry.js";
 import { smoothStream as s, stepCountIs as c, streamText as l } from "ai";
 import { Duration as u, Effect as d, Stream as f } from "effect";
-//#region src/streamText.ts
+// src/streamText.ts
 var p = (e) => {
 	let t = e?.name, n = String(e?.message ?? e ?? "");
 	return t === "AbortError" || t === "TimeoutError" || /\babort(ed)?\b/i.test(n);
@@ -166,5 +166,4 @@ var p = (e) => {
 	};
 	return n(0);
 };
-//#endregion
 export { _ as streamText, x as streamTextWithFallback, b as streamTextWithRetry };

package/dist/streamTranscribe.js

@@ -2,7 +2,7 @@ import { providerFromEnv as e, resolveTranscriptionModel as t } from "./provider
 import { TranscriptionStreamError as n, TranscriptionStreamEvent as r } from "./transcriptionEvent.js";
 import { experimental_streamTranscribe as i } from "ai";
 import { Effect as a, Schema as o, Stream as s } from "effect";
-//#region src/streamTranscribe.ts
+// src/streamTranscribe.ts
 var c = (c) => s.unwrap(a.gen(function* () {
 	let l = yield* a.acquireRelease(a.sync(() => new AbortController()), (e) => a.promise(async () => {
 		e.abort(), c.audio.locked || await c.audio.cancel().catch(() => {});
@@ -70,5 +70,4 @@ var c = (c) => s.unwrap(a.gen(function* () {
 	}));
 	return s.concat(h, g);
 }));
-//#endregion
 export { n as TranscriptionStreamError, r as TranscriptionStreamEvent, c as streamTranscribe };

package/dist/streamWakeups.js

@@ -1,5 +1,5 @@
 import { Effect as e } from "effect";
-//#region src/streamWakeups.ts
+// src/streamWakeups.ts
 var t = /* @__PURE__ */ new WeakMap(), n = (e, n) => (t.get(e)?.get(n)?.readers ?? 0) > 0, r = (n, r) => e.acquireRelease(e.sync(() => {
 	let e = t.get(n);
 	e === void 0 && (e = /* @__PURE__ */ new Map(), t.set(n, e));
@@ -40,5 +40,4 @@ var t = /* @__PURE__ */ new WeakMap(), n = (e, n) => (t.get(e)?.get(n)?.readers
 	};
 	return i;
 };
-//#endregion
 export { i as awaitStreamChange, n as hasStreamReaders, r as watchStream, a as withStreamWakeups };

package/dist/structuredEvidence.js

@@ -1,6 +1,6 @@
 import { normalizeAiCallEvidence as e } from "./callEvidence.js";
 import { wrapLanguageModel as t } from "ai";
-//#region src/structuredEvidence.ts
+// src/structuredEvidence.ts
 var n = (n, r) => {
 	let i = [];
 	return {
@@ -44,5 +44,4 @@ var n = (n, r) => {
 		}
 	};
 };
-//#endregion
 export { n as captureStructuredEvidence };

package/dist/test.js

@@ -1,5 +1,5 @@
 import { MockScript as e, setProviderOverride as t } from "./provider.js";
-//#region src/test/index.ts
+// src/test/index.ts
 var n = (e) => e.turns && e.turns.length > 0 ? e.turns : e.stream && e.stream.length > 0 ? [{ text: e.stream.join("") }] : e.generate ? [{ text: e.generate.text }] : [{ text: "" }], r = (t) => {
 	let r = new e(n(t));
 	return {
@@ -36,5 +36,4 @@ var n = (e) => e.turns && e.turns.length > 0 ? e.turns : e.stream && e.stream.le
 		return t;
 	};
 };
-//#endregion
 export { a as mockAi, o as throwNTimes, i as useMockAi };

package/dist/threads.js

@@ -2,7 +2,7 @@ import { AGENT_MESSAGES_TABLE as e, AGENT_THREADS_TABLE as t } from "./agentCons
 import { AgentThreadAccessDenied as n } from "./agent.js";
 import { Effect as r } from "effect";
 import { boolean as i, id as a, integer as o, json as s, table as c, text as l, timestamp as u } from "@voltro/database";
-//#region src/threads.ts
+// src/threads.ts
 var d = c(t, {
 	id: a({ prefix: "thread" }),
 	tenantId: l().nullable(),
@@ -88,5 +88,4 @@ var d = c(t, {
 	skip: void 0,
 	projection: void 0
 }))[0] ?? null);
-//#endregion
 export { e as AGENT_MESSAGES_TABLE, t as AGENT_THREADS_TABLE, f as agentMessagesTable, d as agentThreadsTable, h as appendMessage, p as createThread, m as ensureThreadOwned, g as getMessages, _ as getThread };

package/dist/tools.js

@@ -1,7 +1,7 @@
 import { jsonSchema as e, tool as t } from "ai";
 import { Context as n, Effect as r, Schema as i } from "effect";
 import { jsonSchemaOf as a } from "@voltro/protocol";
-//#region src/tools.ts
+// src/tools.ts
 var o = (e, t) => e(t), s = Symbol.for("@voltro/ai/Tool"), c = (e) => ({
 	...e,
 	[s]: !0
@@ -64,5 +64,4 @@ var o = (e, t) => e(t), s = Symbol.for("@voltro/ai/Tool"), c = (e) => ({
 	}
 	return o;
 };
-//#endregion
 export { g as buildSdkTools, c as defineTool, l as isTool, u as resolveToolModuleHandlers };

package/dist/transcribe.js

@@ -3,7 +3,7 @@ import { aiCallAttributes as n, withAiCallSpan as r } from "./observe.js";
 import { aiGenerationError as i } from "./api.js";
 import { experimental_transcribe as a } from "ai";
 import { Effect as o } from "effect";
-//#region src/transcribe.ts
+// src/transcribe.ts
 var s = async (e, t) => {
 	if ("bytes" in e) return {
 		audio: e.bytes,
@@ -60,5 +60,4 @@ var s = async (e, t) => {
 		catch: (e) => i(e)
 	}));
 };
-//#endregion
 export { l as transcribe };

package/dist/transcriptionEvent.js

@@ -1,5 +1,5 @@
 import { Schema as e } from "effect";
-//#region src/transcriptionEvent.ts
+// src/transcriptionEvent.ts
 var t = class extends e.TaggedError()("TranscriptionStreamError", {
 	reason: e.Literals([
 		"input",
@@ -42,5 +42,4 @@ var t = class extends e.TaggedError()("TranscriptionStreamError", {
 		durationSec: e.optional(e.Number)
 	})
 ]);
-//#endregion
 export { t as TranscriptionStreamError, i as TranscriptionStreamEvent };

package/dist/workflow.js

@@ -8,7 +8,7 @@ import { AI_INFERENCES_TABLE as f, inferenceSignalName as p, offloadedInferenceI
 import { Effect as h, Schema as g } from "effect";
 import { eq as _ } from "@voltro/database";
 import { BudgetHoldExpired as v, CurrentWorkflowExecutionId as y, SuspensionUnavailable as b, WorkflowCannotPark as x, awaitSignalSuspending as S, step as C, suspendForBudget as w } from "@voltro/workflow";
-//#region src/inferStep.ts
+// src/inferStep.ts
 var T = {
 	maxAttempts: 4,
 	strategy: "exponential",
@@ -307,5 +307,4 @@ var T = {
 		})).pipe(h.tap((e) => j(t, n, "generateObject", e, r.rendered)), h.tap(() => M(t, r.rendered)), h.map((e) => e.object))
 	}));
 };
-//#endregion
 export { T as AI_STEP_RETRY, z as AiBudgetMisuseError, F as AiOffloadError, I as AiOffloadMisuseError, D as DEFAULT_AI_BUDGET_NAME, E as DEFAULT_OFFLOAD_TIMEOUT_MS, U as TranscriptionStepResult, K as aiObjectStep, H as aiStep, G as mediaStep, u as promptDigest, O as promptRecordFor, k as resolveStepPrompt, W as transcribeStep };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@voltro/ai",
-  "version": "0.86.0",
+  "version": "0.88.0",
   "description": "Effect-native AI primitives — generateText / generateObject / streamText, tools, agents, media generation, cost tracking + budgets, and provider fallback.",
   "keywords": [
     "voltro",
@@ -53,21 +53,22 @@
     "node": ">=24.0.0"
   },
   "dependencies": {
-    "@ai-sdk/anthropic": "^4.0.56",
-    "@ai-sdk/openai": "^4.0.69",
-    "@voltro/cache": "0.86.0",
-    "@voltro/database": "0.86.0",
-    "@voltro/protocol": "0.86.0",
-    "@voltro/workflow": "0.86.0",
+    "@ai-sdk/anthropic": "^4.0.91",
+    "@ai-sdk/openai": "^4.0.104",
+    "@voltro/cache": "0.88.0",
+    "@voltro/database": "0.88.0",
+    "@voltro/env": "0.88.0",
+    "@voltro/protocol": "0.88.0",
+    "@voltro/workflow": "0.88.0",
     "ajv": "^8.20.0",
     "ajv-formats": "^3.0.1",
-    "ws": "^8.21.3"
+    "ws": "^8.22.0"
   },
   "peerDependencies": {
-    "@ai-sdk/gateway": "^4.0.74",
-    "@ai-sdk/provider": "^4.0.10",
-    "ai": "^7.0.92",
-    "effect": "4.0.1"
+    "@ai-sdk/gateway": "^4.0.85",
+    "@ai-sdk/provider": "^4.0.17",
+    "ai": "^7.0.105",
+    "effect": "4.1.0"
   },
   "publishConfig": {
     "access": "public"