Fireworks prompt caching is enabled by default (automatic prefix matching),
but on serverless infrastructure, requests hit random replicas. Without
session affinity, the per-replica cache misses, negating cache hit rates
and the discounted cacheRead pricing.
Changes:
- Add sendSessionAffinityHeaders and supportsCacheControlOnTools
to AnthropicMessagesCompat interface
- Send x-session-affinity header for Fireworks (and Cloudflare AI
Gateway Anthropic) when sessionId is available and caching is enabled
- Omit cache_control on tool definitions for Fireworks (unsupported
per https://docs.fireworks.ai/tools-sdks/anthropic-compatibility)
- Default supportsEagerToolInputStreaming to false for Fireworks
(unsupported field)
- Default supportsLongCacheRetention to false for Fireworks
(cache_control.ttl not supported)
- Add compat settings to Fireworks models in generate-models.ts
- Update generated models with Fireworks compat settings
- Add integration tests for session affinity and tool compat
Refs: https://docs.fireworks.ai/guides/prompt-caching
Refs: https://docs.fireworks.ai/tools-sdks/anthropic-compatibility
DashScope / Aliyun Qwen (OpenAI-compatible) rejects `tools: []`
with HTTP 400 `"[] is too short - 'tools'"`. Five providers used
a truthy check (`if (context.tools)`) that treated an empty array
as "send tools", so `pi --no-tools` produced `tools: []` in the
request body. Matching the Google provider's pattern, we now
guard on `context.tools.length > 0`:
- openai-completions.ts
- openai-responses.ts
- openai-codex-responses.ts
- azure-openai-responses.ts
- anthropic.ts
The openai-completions fallback that emits `tools: []` when the
conversation has tool history (required by LiteLLM / Anthropic
proxies) is preserved via the existing `else if (hasToolHistory)`
branch.
closes#3649
Co-authored-by: 槐聚 <huaiju@zbyte-inc.com>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Mario Zechner <badlogicgames@gmail.com>
Replace the deprecated global anthropic-beta header with per-tool
eager_input_streaming: true. Fix empty anthropic-beta header causing
400 errors on models with no beta features.
fixes#3175
Bypass the Anthropic SDK streaming parser entirely. Use
client.messages.create().asResponse() and decode the SSE stream
ourselves with defensive JSON parsing that repairs invalid escape
sequences and control characters inside string literals.
- Switch from SDK .stream() to .asResponse() + pi-owned SSE decoder
- Add repairJson() / parseJsonWithRepair() to json-parse.ts
- Add anthropic-sse-parsing.test.ts regression for malformed tool deltas
- Update github-copilot-anthropic.test.ts mock to match new call path
- Update deprecated claude-3-5-haiku-20241022 refs to claude-haiku-4-5
- Remove stale non-reasoning model test
fixes#3175
Exposes the new ThinkingConfig.display field on Anthropic and Bedrock
Claude providers. Defaults to 'summarized' so Claude Opus 4.7 and Mythos
Preview keep returning thinking text despite Anthropic's silent default
change to 'omitted'. Set to 'omitted' explicitly to skip thinking
streaming for faster time-to-first-text-token.
Mutate persisted tool-call blocks in place on function_call completion,
remove partialJson, and emit the same reference on toolcall_end.
Add regression coverage for persisted block cleanup and event identity.
fixes#3078
Add optional `client` field to AnthropicOptions so callers can pass
alternative SDK clients (e.g. AnthropicVertex) that
share the same messages.stream() API. When omitted, behavior is unchanged.
Signed-off-by: sallyom <somalley@redhat.com>
- Map redacted_thinking to ThinkingContent with redacted: true instead of
adding a new content type. The opaque payload goes in thinkingSignature,
thinking text is set to "[Reasoning redacted]" so it renders naturally
everywhere. Cross-model transform drops redacted blocks.
- Skip interleaved-thinking-2025-05-14 beta header for Opus 4.6 / Sonnet 4.6
where adaptive thinking makes it deprecated/redundant.
- Do not send temperature when thinkingEnabled is true (incompatible with
both adaptive and budget-based thinking).
Based on #1665 by @tctev
Add metadata?: Record<string, unknown> to StreamOptions so providers
can extract fields they understand. Anthropic provider extracts user_id
for abuse tracking and rate limiting. Other providers ignore it.
Based on #1384 by @7Sageer, reworked to use a generic type instead of
Anthropic-specific typing on the base interface.
- Add adaptive thinking mode (type: 'adaptive') for Opus 4.6+
- Add effort parameter ('low', 'medium', 'high', 'max') for adaptive thinking
- thinkingEnabled now auto-detects: adaptive for 4.6+, budget-based for older
- streamSimple/completeSimple map ThinkingLevel to effort levels for Opus 4.6
- Add tests for Opus 4.6 adaptive thinking and GPT-5.3 Codex
- Update @anthropic-ai/sdk to 0.73.0
- Update @aws-sdk/client-bedrock-runtime to 3.983.0
- Update @google/genai to 1.40.0
- Remove fast-xml-parser override (no longer needed)
