agent-accelerator 0.5.8 → 0.6.0

package/README.md

@@ -126,7 +126,7 @@ An `Agent` holds:
 | `cache`           | `CacheOption`                                        | `"implicit" \| "5m" \| "30m" \| "1h" \| "12h"`. Prompt-cache lifetime; unsupported values warn and fall back (see [Cache](#cache)).                                                                             |
 | `serviceTier`     | `"flex" \| "priority"`                               | Cost / priority routing where supported. Omit for standard routing.                                                                                        |
 | `maxTurns`        | `number`                                             | Maximum model → tool → model loops per `run`. Infinite by default (`0` or omitted = no limit). Hitting a finite limit with pending tool calls warns and sets `finishReason: "max_turns"`. |
-| `limits`          | `AgentRunLimits`                                     | `{ maxTurns?, deadlineMs?, maxToolCallsPerRun?, maxTotalTokens?, requestTimeoutMs?, firstByteTimeoutMs?, streamIdleTimeoutMs? }`. The first four are unlimited by default. The three network timeouts default to 300 s / 120 s / 300 s so a silent connection fails instead of hanging; set one to `0` or `Infinity` to turn it off. A timeout raises a retryable `TimeoutError`. See [Safety and operations](docs/guides/safety-and-ops.md#network-timeouts). |
+| `limits`          | `AgentRunLimits`                                     | `{ maxTurns?, deadlineMs?, maxToolCallsPerRun?, maxTotalTokens?, requestTimeoutMs?, firstByteTimeoutMs?, streamIdleTimeoutMs? }`. The first four are unlimited by default; hitting `maxTurns` ends the run with `finishReason: "max_turns"`, the other three throw a `BudgetExceededError`. The three network timeouts default to 300 s / 120 s / 300 s so a silent connection fails instead of hanging; set one to `0` or `Infinity` to turn it off. A timeout raises a retryable `TimeoutError`. See [Safety and operations](docs/guides/safety-and-ops.md#network-timeouts). |
 | `budget`          | `Budget`                                             | `{ maxCostUsd?, maxTotalTokens?, maxDurationMs?, maxTotalSubagents? }`, per run, workers included. Unlimited by default; exceeding it throws `BudgetExceededError` with partial usage. |
 | `sessionId`       | `string`                                             | Stable identifier used for cache affinity. Auto-generated when omitted.                                                                                    |
 | `headers`         | `Record<string,string>`                              | Additional headers merged into every request.                                                                                                              |
@@ -720,7 +720,7 @@ It also:
 * maps `tool_choice` (`auto` omitted; `none`, `required`, and function pins supported);
 * treats an error chunk inside a stream (HTTP 200) like an HTTP error: its status comes from the chunk's `code`, or from its `error_type` (`rate_limit_exceeded` 429, `provider_overloaded` 503, `provider_unavailable` 502, `timeout` 504, `server` 500), so `runRetries` retries the retryable ones;
 * reports `cached_tokens` / `cache_write_tokens` / reasoning tokens, preferring provider-reported `cost` when present (a reported `0`, a free model or a cache hit, stays a known `$0`);
-* accepts `anthropic/...` and `~anthropic/...` model names for the `5m` / `1h` cache setting, and sends at most 4 `stop` sequences (the Chat Completions maximum; extra ones are dropped with one `provider_warning`).
+* accepts `anthropic/...` and `~anthropic/...` model names for the `5m` / `1h` cache setting.
 
 `serviceTier` values other than `flex` / `priority` are omitted (standard routing).
 
@@ -1089,16 +1089,35 @@ Credential headers are always redacted (`Authorization`, `x-goog-api-key`, anyth
 ---
 ## Errors
 
-Provider failures throw `AgentAccelProviderError` — one actionable line,
+Provider failures throw a `ProviderError` — one actionable line,
 never a wire dump:
 
 ```text
 [openrouter/nvidia/nemotron-3.5-lightning:free] request failed (404): No endpoints found that support input video
 ```
 
-Raw details (`statusCode`, `url`, truncated body) stay attached as
+Branch on the class and `err.retryable` (true for HTTP 408/425/429/5xx,
+network errors and a cut stream), never on the message:
+
+```ts
+import { ProviderError, TimeoutError } from "agent-accelerator";
+
+try {
+  await agent.run("...");
+} catch (err) {
+  if (err instanceof ProviderError && err.retryable) retryLater(); // e.g. a 429 or 503
+  else if (err instanceof TimeoutError) retryLater(); // a request / first-byte / idle timeout
+  else throw err;
+}
+```
+
+`err.code` is the SDK kind (`"provider"`, `"incomplete_stream"` for a cut
+stream; `"timeout"`, `"config"`, ... on the other classes); the provider's own
+error code (e.g. `"not_found"`, `"rate_limit_exceeded"`) is `err.providerCode`.
+Raw details (`statusCode`, `url`, truncated body, `providerCode`) stay attached as
 non-enumerable properties: available programmatically, invisible in runtime
-dumps. URL query strings are stripped (keys sometimes live there); headers
+dumps. Input a model cannot take (a modality the catalog says it lacks, video on
+the OpenAI Responses wire) fails before the network with a `ValidationError`. URL query strings are stripped (keys sometimes live there); headers
 and cookies are never attached. Helpers:
 
 * `toConciseProviderError(err, providerId, modelId)` — collapse any provider error.
@@ -1433,6 +1452,8 @@ saveSessionFile(".session.json", agent, telemetry);
 
 ```bash
 bun run typecheck     # tsc --noEmit
+bun run lint          # Biome: unused code is an error, `any` a warning
+bun run format        # Biome formatter (120 columns); `biome format src examples` only checks
 bun test              # bun test test/
 bun run update-models # refresh model catalog cache (supports --force, --ttl=24h)
 ```

package/examples/prompts/SYSTEM_PROMPT.md

@@ -1,12 +1,14 @@
 You are **Agent Accelerator**, an expert, high-performance AI assistant.
 
 ### Operational Directives
+
 - **Direct Execution**: Be concise, direct, and factual. Use structured Markdown (headings, lists, tables) for clarity.
 - **Thinking Traces**: When reasoning is enabled, think methodically before concluding. Keep final answers synthesized and free of raw thinking artifacts.
 - **File Discipline**: Always inspect/read target files before editing; perform surgical, minimal modifications. State absolute or workspace-relative paths clearly.
 - **Code & Shell Quality**: Prefer idiomatic, type-safe solutions. Avoid destructive operations unless explicitly requested.
 
 ### Tool-Calling Protocols
+
 - **Strict Schema Conformance**: Adhere strictly to the tool's parameter schema. Never invent, hallucinate, or pass unrecognized arguments.
 - **Parallel Dispatch**: When multiple tool calls are independent, batch and execute them simultaneously in a single turn.
 - **No Narration**: Invoke tools immediately. Do not announce or narrate tool invocations beforehand (e.g., avoid "I will now read...").

package/examples/prompts/SYSTEM_PROMPT_AGENT.md

@@ -1,6 +1,7 @@
 You are an Agent — you evaluate user objectives and decide whether to solve them directly or delegate to specialized parallel workers with the `subagents` tool.
 
 ### Workflow & Decision Matrix
+
 1. **Triage**:
    - Single-focus, factual, code-generation, or conversational prompts → Answer directly, concisely, and completely.
    - Do NOT start workers for trivial, single-step tasks.
@@ -13,10 +14,12 @@ You are an Agent — you evaluate user objectives and decide whether to solve th
      - `task`: Focused, non-overlapping research query or execution directive.
 3. **Aggregate**:
    - Each worker answers with a header line and its findings:
+
    ```
    [worker market-analyst-1a2b3c4d] replied:
    Findings from the worker
    ```
+
    - Use `command: "message"` with the worker's `id` to ask a follow-up; the worker remembers the conversation. `command: "close"` when you no longer need it.
 4. **Synthesize**:
    - Deliver an editorial-grade, cohesive report that integrates all worker insights.
@@ -24,5 +27,6 @@ You are an Agent — you evaluate user objectives and decide whether to solve th
    - Structure with executive summaries, comparative Markdown tables, and concrete strategic conclusions. Never present an uncurated list of worker summaries.
 
 ### Execution Rules
+
 - **Tool Protocol**: Call `subagents` directly without conversational preambles. Start parallel workers with several `spawn` calls in one turn.
 - **Context Isolation**: Workers start blank; provide complete background in their `instructions` and `task`.

package/examples/prompts/SYSTEM_PROMPT_TOOLS.md

@@ -1,6 +1,7 @@
 You are a precise, tool-equipped AI agent engineered for accurate execution and high-efficiency operations.
 
 ### Tool Invocation Protocols
+
 - **Intent Verification**: Evaluate whether tool execution is necessary. If the query is conversational or already answerable with existing context, answer directly.
 - **Strict Schema Adherence**: Validate all arguments against the declared JSON schema before calling. Never invent parameters, omit required properties, or pass invalid types.
 - **Parallel Dispatch**: Whenever multiple tool operations are independent (e.g., multi-file reads, batch queries, concurrent searches), invoke all tool calls concurrently in a single turn.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-accelerator",
-  "version": "0.5.8",
+  "version": "0.6.0",
   "description": "High-performance SDK for AI agents with a unified API across AI providers. Optimized for high cache hit rates, observability, and seamless agent orchestration. Built to power production-grade agentic systems and harnesses for teams at any scale.",
   "license": "MIT",
   "author": "Sashvat Bharat",
@@ -41,6 +41,8 @@
   "type": "module",
   "scripts": {
     "typecheck": "tsc --noEmit",
+    "lint": "biome lint src examples",
+    "format": "biome format --write src examples",
     "test": "bun test test/",
     "update-models": "bun src/update-models.ts",
     "fetch-latest-provider-docs": "bun references/documentations/fetch-latest.ts"
@@ -49,6 +51,7 @@
     "zod": "^4.4.3"
   },
   "devDependencies": {
+    "@biomejs/biome": "2.6.6",
     "@types/bun": "latest",
     "typescript": "^7.0.0",
     "zod": "^4.4.3"