Changelog
All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
0.57.1 (2026-11-13)
Bug Fixes
-
anthropic_vertex: download URL files and add anthropic_vertex to the acceptance suite (#585) (1aef911)
0.57.0 (2026-11-13)
Features
-
accept symbol param types and deprecate Ruby class types (#572) (3aa722c)
-
agent: add deferred tools that a tool can load partway through a run (#557) (6ee3acb)
-
anthropic: continue pause_turn turns automatically (#539) (b6fbedc)
-
anthropic: load deferred tools natively on anthropic and anthropic_vertex (#575) (bba6541)
-
evals: deprecate
Riffer::Evalsand config.evals (#573) (c66a0ba) -
openai: load deferred tools natively through additional_tools (#576) (d57837f)
-
providers: load deferred tools on any provider through a gateway tool (#578) (92444fc)
Bug Fixes
0.56.0 (2026-11-09)
⚠ BREAKING CHANGES
-
messages:
Assistant#to_hwritespartsinstead ofcontent,tool_callsandreasoning, andto_hno longer acceptsparts:. Build the old fields from content, tool_calls and reasoning on the message. This also changes the assistant entries inRiffer::Providers::Mock#callsandRiffer::Evals::ScenarioResult#to_h.
Features
-
add
Riffer::Deprecationfor deprecation warnings (#549) (6ad6046) -
add
Riffer.config.future for opt-in future flags (#556) (6e49471) -
messages: keep assistant turns in order with Assistant#parts (#527) (15e8eb9)
Bug Fixes
0.55.0 (2026-10-29)
⚠ BREAKING CHANGES
-
gemini: Gemini and Google Cloud structured output schemas are strict.
-
anthropic: the anthropic gem is no longer used. config.anthropic.client takes a
Riffer::Providers::Anthropic::Client; an Anthropic::Client raises a migration error. SDK error classes are replaced byRiffer::APIErrorand Faraday::Error. There are no default retries (the SDK retried twice); add them through connection:. ANTHROPIC_AUTH_TOKEN and ANTHROPIC_CUSTOM_HEADERS are no longer read. Streamed ToolCallDelta events carry the real item_id and name. -
gemini: gemini-2.x models ignore responseFormat on the Developer API and return prose, so structured output needs gemini-3 or later.
-
gemini: Gemini::Client no longer accepts open_timeout, read_timeout, write_timeout, proxy_address or proxy_port; pass a connection: instead. The default read timeout is 600s (was 60s). Non-success responses raise
Riffer::APIError(still aRiffer::Error), and transport failures are Faraday::Errors rather than Net::* errors. A missing API key raisesRiffer::ArgumentError(an ::ArgumentError, not aRiffer::Error) instead of a 4xxRiffer::Error, and GEMINI_API_KEY is now read when config.gemini.api_key is unset.
Features
-
add
Riffer::HTTP::ClientFaraday transport andRiffer::APIError(#503) (9ee2dfc) -
anthropic: replace the anthropic gem with a hand-rolled Messages API client (#506) (6388c44)
-
gemini: add web search through Google Search grounding (#510) (ee3c820)
-
gemini: move Gemini::Client onto
Riffer::HTTP::Client(#505) (1a6798c) -
gemini: send structured output as responseFormat and replay thought signatures (#507) (a5299ef)
-
params: add primitive variants to
Riffer::Params::AnyOf(#508) (98483b4) -
providers: add anthropic_vertex provider for Claude on Agent Platform (#512) (10a55a1)
-
providers: add google_cloud provider for Gemini on Agent Platform (#511) (b0f44d6)
Bug Fixes
-
amazon_bedrock: send valid document names and “{}” for streamed calls without arguments (#522) (992f678)
-
gemini: send parallel tool results back in one turn (#518) (79d11ea)
-
gemini: strict structured output schema, placeholder thought signatures, one TokenUsageDone per stream (#520) (a3a1063)
0.54.0 (2026-10-20)
Features
-
params: add
Riffer::Params::AnyOf(#491) (65493ca) -
params: add arrays of arrays with nested
of:(#498) (2b3449c) -
structured_output: append prompted schema to user messages (#492) (332db21)
Bug Fixes
0.53.0 (2026-10-13)
Features
0.52.1 (2026-10-09)
Bug Fixes
-
structured_output: return the extracted JSON as response content (#484) (cbea668)
0.52.0 (2026-10-08)
Features
-
structured_output: add experimental prompted strategy (#477) (f517fe1)
0.51.0 (2026-10-07)
Features
-
structured_output: recover objects wrapped in code fences or prose (#473) (59a6d40)
Bug Fixes
0.50.1 (2026-10-02)
Bug Fixes
0.50.0 (2026-09-28)
Features
-
anthropic: capture and replay thinking blocks (#463) (4a25851)
-
open_ai: capture and replay reasoning items (#464) (03d54a9)
Bug Fixes
0.49.0 (2026-09-27)
⚠ BREAKING CHANGES
-
config: provider string settings must be a String or nil; evals.judge_model must be in “provider/model” form; mcp.credentials must respond to call; mcp.discovery_runner must be a
Riffer::Runner. Struct behaviour ([] / to_a / members / value equality) is gone from those sections. Riffer::Tracing.reset! is removed; the tracing backend is read from config on every call instead of being cached. -
agent:
Riffer.config.experimental_history_healing is removed, as are Response#healed_tool_call_ids and StreamEvents::Interrupt#healed_tool_call_ids (and its to_h key). Orphan pruning on Agent.new is now always on. Interrupts never fill placeholders; call agent.session.discard_pending_tool_calls instead and use its return value for the filled call ids.
Features
-
agent: always repair seeded sessions, add discard_pending_tool_calls (#456) (b1e8933)
-
amazon_bedrock: capture and replay reasoning content (#459) (5048a8c)
-
open_router: capture and replay reasoning_details (#453) (a3e2523)
Code Refactoring
-
config: split
Riffer::Configinto one file per section (#458) (c493684)
0.48.0 (2026-09-23)
⚠ BREAKING CHANGES
-
messages: Riffer::Messages::FilePart is renamed to
Riffer::Messages::User::FilePartwith no alias. ToolCall no longer responds to Struct-only methods such as [] or to_a; use its readers instead.
Features
-
identify agents and evals with default tags (#452) (22ab54c)
-
messages: persist reasoning on assistant messages (#446) (7fc64ab)
Bug Fixes
Code Refactoring
0.47.2 (2026-09-21)
Bug Fixes
-
providers: raise IncompleteStreamError when a stream ends without its terminal event (#445) (d142e23)
-
stop relying on the transitive cgi gem for CGI.escapeHTML (#443) (821f3e3)
0.47.1 (2026-09-11)
Bug Fixes
-
params: keep optional Hash params nullable in strict JSON Schema (#433) (c7308fd)
-
params: raise for bare Hash and Array params under strict schemas (#434) (174ad92)
0.47.0 (2026-09-11)
Features
0.46.1 (2026-09-11)
Bug Fixes
0.46.0 (2026-09-08)
⚠ BREAKING CHANGES
-
Response#blocked?, Response#interrupted?, and Response#interrupt_reason are removed. Check response.outcome.reason for :guardrail_blocked, :interrupted, or :max_steps instead; response.tripwire still carries the full tripwire object. Response.new now requires outcome:.
Features
Bug Fixes
-
anthropic: merge caller output_config with structured output (#422) (e1a60e6)
-
runner: make Fibers runner safe inside a running Async reactor (#424) (47d3706)
0.45.0 (2026-09-03)
Features
0.44.0 (2026-08-28)
Features
-
file attachment downloading for user messages (#407) (4e78868)
-
registry liveness, explicit registration, and
Riffer::Testingstub factories (#415) (6935749)
0.43.0 (2026-08-27)
⚠ BREAKING CHANGES
-
call_with_validation no longer raises
Riffer::ValidationError/Riffer::TimeoutError/Riffer::Error— it always returns aRiffer::Tools::Response.Riffer::TimeoutErrorhas been removed; delete anyrescue Riffer::TimeoutErrorand checkresponse.error_typeinstead. Custom runtimes overriding dispatch_tool_call no longer inherit rescues from the base class.
Features
0.42.0 (2026-08-25)
⚠ BREAKING CHANGES
-
duplicate identifiers among named direct subclasses of
Riffer::ToolorRiffer::Agentnow raiseRiffer::DuplicateIdentifierErroron lookup. Agent.find/.all (and the new Tool.find/.all) cover named direct subclasses only - anonymous classes are excluded even with an explicit identifier. Riffer::Helpers::ClassNameConverter is removed; useRiffer::Helpers::Identifier.derive(the separator: keyword is dropped).
Features
0.41.0 (2026-08-14)
⚠ BREAKING CHANGES
-
Inference providers no longer support extra configuration parameters (e.g. timeouts). Instead, a provider client object can be defined in the riffer config directly. Credential and endpoint parameters are still supported.
Features
0.40.0 (2026-08-12)
Features
0.39.0 (2026-07-29)
Features
0.38.1 (2026-07-29)
Performance Improvements
0.38.0 (2026-07-21)
Features
0.37.1 (2026-07-13)
Bug Fixes
0.37.0 (2026-07-10)
Features
0.36.0 (2026-07-03)
⚠ BREAKING CHANGES
-
config.metrics, the Riffer::Metrics module, and Riffer::Metrics::Otel are removed. Hosts wiring a metrics backend must drop that configuration.
Features
0.35.0 (2026-06-25)
Features
0.34.0 (2026-06-22)
⚠ BREAKING CHANGES
-
telemetry:
config.tracing.tracer_providerandconfig.metrics.meter_providerare removed. Assign the OTEL backend instead:config.tracing.backend = Riffer::Tracing::Otel.build(passprovider:to override the global provider).
Features
-
emit a riffer.guardrail.duration metric for guardrail execution (#338) (7b0dfb0)
-
telemetry: pluggable tracing/metrics backend (OTEL opt-in) (#329) (d647f54)
0.33.0 (2026-06-18)
⚠ BREAKING CHANGES
-
reported input_tokens grows by the cache token counts on Anthropic and Bedrock, and output_tokens grows by the thinking token count on Gemini, whenever those features are active.
Features
-
add the Riffer::Metrics OpenTelemetry port foundation (#325) (92d060c)
-
add tracing foundation with optional OTEL backend (#307) (938194c)
-
emit a chat span per LLM call with normalized finish reasons (#312) (42f39dd)
-
emit an execute_guardrail span per guardrail (#324) (b280c71)
-
emit gen_ai.client.operation.duration metric (#326) (a9399b8)
-
emit riffer.gen_ai.cost metric from TokenUsage cost (#328) (1ee4772)
-
support user-explicit skill activation and dedupe re-activations (#305) (f95b908)
Code Refactoring
0.32.1 (2026-06-10)
Bug Fixes
0.32.0 (2026-06-08)
⚠ BREAKING CHANGES
-
TokenUsage#cache_creation_tokensis renamed tocache_write_tokens(also reflected inTokenUsage#to_h). Update any code reading that attribute or hash key. -
apply module/class conventions consistently (#298)
Features
-
add Bedrock prompt caching and surface cached tokens (#300) (407390b)
-
add progressive discovery for MCP tools by default (0a37485)
Code Refactoring
0.31.0 (2026-06-03)
Features
0.30.0 (2026-06-03)
⚠ BREAKING CHANGES
-
unlimited max_steps is now represented as
nilat the agent level (set viamax_steps nil), not Float::INFINITY —Riffer::Agent::Config#max_stepsmay be nil and is no longer normalized.Riffer::Paramsparameter types are typed Module (widened from Class) to honestly includeRiffer::Params::Boolean, which is a Module.
Features
-
Riffer::Agent::Serializerfor transferable agent definitions (#293) (99134b0)
0.29.1 (2026-06-01)
Bug Fixes
0.29.0 (2026-05-29)
⚠ BREAKING CHANGES
-
Public API reshape on
Riffer::Agent. Downstream consumers must migrate the following surfaces:
Features
Code Refactoring
-
extract
Riffer::Agent::Run, eagerly resolve per-agent state (#268) (1d9e141) -
move leaf types out of root
Riffernamespace (#282) (3637a53)
0.28.0 (2026-05-08)
⚠ BREAKING CHANGES
-
Custom subclasses of Riffer::ToolRuntime that override around_tool_call or dispatch_tool_call must accept the new assistant_message: kwarg (or **kwargs). Existing overrides that omit it will raise ArgumentError: unknown keyword: :assistant_message.
Features
-
add history mutation API to
Riffer::Agent(#249) (d980daa) -
forward assistant_message to ToolRuntime hooks (#247) (3d5f935)
0.27.2 (2026-05-04)
Bug Fixes
0.27.1 (2026-05-04)
Bug Fixes
0.27.0 (2026-05-01)
Features
-
integrate MCP server tools into agent tool resolution (6c7a09a)
Bug Fixes
0.26.0 (2026-04-29)
Features
0.25.0 (2026-04-29)
⚠ BREAKING CHANGES
-
Removed Riffer::Skills::Adapter#activate_tool. Set the activation tool via
Riffer.config.skills.default_activate_tool, or per-agent withskills do; activate_tool MyTool; end. -
Riffer::Skills::Adapter.newnow requires skill_activate_tool:. The agent wires this up; only matters if you construct an adapter yourself. Custom adapters that override initialize must call super.
Features
0.24.2 (2026-04-23)
Bug Fixes
0.24.1 (2026-04-23)
Bug Fixes
0.24.0 (2026-04-20)
Features
0.23.0 (2026-04-15)
Features
0.22.0 (2026-04-09)
⚠ BREAKING CHANGES
-
Tool.namereturnsnamespace/tool_nameinstead ofnamespace__tool_name.
Features
Bug Fixes
0.21.0 (2026-04-09)
Features
-
amazon_bedrock: support S3 URI file sources (#190) (243e5b1)
-
normalize consecutive same-role messages before provider serialization (#201) (1ac986d)
0.20.0 (2026-03-26)
Features
Bug Fixes
0.19.0 (2026-03-25)
Features
0.18.0 (2026-03-13)
Features
0.17.0 (2026-03-06)
⚠ BREAKING CHANGES
-
rename
tool_contexttocontext(#159)
Features
-
add experimental ToolRuntime abstraction for tool execution (#156) (0ca7563)
-
add interrupt! public method for clean loop interrupts (#155) (a4cc877)
-
auto-derive step offset on resume for max_steps enforcement (#154) (fb97dbe)
Bug Fixes
Code Refactoring
0.16.1 (2026-03-03)
Bug Fixes
0.16.0 (2026-02-27)
Features
-
add Riffer::Boolean sentinel type for boolean params (#147) (5337cf3)
-
nested params DSL and strict schema for all providers (#144) (2984855)
0.15.1 (2026-02-25)
Bug Fixes
0.15.0 (2026-02-25)
Features
0.14.0 (2026-02-24)
Features
-
add file attachment support for user messages (#129) (0d08eb1)
-
add web_search option for OpenAI and Anthropic providers (#126) (7e7e793)
-
replace eval prompt API with semantic fields (#132) (5d99d5a)
-
replace Profile/Metric eval system with EvaluatorRunner (#138) (ebf2696)
Bug Fixes
-
add –comment flag to claude code review prompt (#122) (534bd59)
-
add RBS annotations to web search constants and fix test provider docs (#130) (7cebe05)
-
correct Amazon Bedrock model options to use Converse API structure (#133) (3af4562)
-
widen run_eval input type to accept String or messages array (#124) (f96214c)
0.13.0 (2026-02-12)
Features
0.12.0 (2026-02-11)
⚠ BREAKING CHANGES
-
Agent#generate now returns
Riffer::Agent::Responseinstead of String. Use response.content or response.to_s for the text.
Features
-
add evals primitive for LLM-as-judge evaluations (#101) (8fd7b36)
-
add guardrails primitive for input/output processing (#100) (48d8bad)
-
add inline RBS type annotations with Steep type checking (#103) (02ae559)
Bug Fixes
Miscellaneous Chores
0.11.0 (2026-02-04)
Features
-
add class methods for generate and stream to
Riffer::Agent(#97) (597636a) -
Add token usage tracking in
Riffer::Agent(#102) (6044914)
0.10.0 (2026-01-30)
Features
Bug Fixes
0.9.0 (2026-01-28)
Features
-
implement
Riffer::Tools::Responsefor consistent tool result handling (#91) (df44f1f)
0.8.0 (2026-01-26)
Features
0.7.0 (2026-01-21)
Features
0.6.1 (2026-01-16)
Bug Fixes
0.6.0 (2026-01-14)
Features
0.5.1 (2026-01-10)
Bug Fixes
-
update gem details to reflect janeapp ownership (#66) (06a008d)
-
update GitHub Pages deployment action in release workflow (#68) (e2f7961)
0.5.0 (2026-01-06)
Features
0.4.2 (2025-12-29)
Bug Fixes
0.4.1 (2025-12-29)
Bug Fixes
-
add conditional check for docs job execution based on release creation (#58) (97bc6f7)
-
suppress output during documentation generation (#53) (6b7f2d9)
-
update rdoc command to target the lib directory (#56) (c319efe)
0.4.0 (2025-12-29)
Features
0.3.2 (2025-12-29)
Bug Fixes
0.3.1 (2025-12-29)
Bug Fixes
0.3.0 (2025-12-29)
Features
Bug Fixes
-
remove manifest file from release configuration (#41) (2f898d8)
-
remove release-type configuration from release workflow (#42) (e270a6c)
-
restructure release configuration and update manifest format (#45) (d07694c)
-
simplify release configuration by removing unnecessary package structure (#40) (8472967)
0.2.0 (2025-12-28)
Features
[0.1.0] - 2024-12-20
Added
-
Core Framework: Foundation for building AI-powered applications and agents
-
Configuration System: Global and instance-level configuration management
-
Agents: Base agent class for building conversational agents
-
Messages: Complete message system with support for User, Assistant, System, and Tool messages
-
Providers: Pluggable provider architecture
-
OpenAI Provider: Full integration with OpenAI API for text generation and streaming
-
Test Provider: Built-in test provider for development and testing
-
Stream Events: Streaming response support with TextDelta and TextDone events
-
Zeitwerk Autoloading: Modern Ruby autoloading for clean code organization
-
Comprehensive Test Suite: Full RSpec test coverage with VCR cassettes for API mocking
-
StandardRB Code Style: Enforced code formatting and linting