llm.rb 8.1.0 → 10.0.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 8aa3ee461642fb157bece63a4ebe00ceda8ec66ce24df5c842efdcc176861a53
-  data.tar.gz: 2d26e36b812704a80e5c8ba4814cfbec770afd5694be71b69d7937422f9a642c
+  metadata.gz: 6ba756238fa72e58ba774567a0c8e2a6d7351cb6313f9c8c08cbdeec8ec9cfa4
+  data.tar.gz: cba8295670dab2843cec902ae97b7ae14e775359380ba401ca5a0066eb60ad0e
 SHA512:
-  metadata.gz: 3a30bf9d5309bf49c660137ed5e81b74f9b028f8846077f3db0b7c92745a5d96b16115db765a4bd1970ba0cbaaa7bd805e0a4a37c04c7e63aacdf3d019d268ec
-  data.tar.gz: 4e297d159dc459ee9ec228862f271b7a21be48ce06f092773c4b56d9cc007252b1cfeb66a119c7e14f3e683213e5923d34b0c256397b92cfa981cd47fe023008
+  metadata.gz: b8347b2adfe05a4700ec42e0ed5992a1332355bd20330590d8b3de214d980476a490855ff7e69b5b36c75f3684304c4ee61bdff9ecbcf8001f0b477b8010d064
+  data.tar.gz: a41512ffbc52b3665118161251441152389ca9daba1a6f4e010303490938dc33393da62f5e821521b2a9f4b45d85fd219b558fa7d2e185c24f43777d26e36a14

data/CHANGELOG.md

@@ -2,6 +2,197 @@
 
 ## Unreleased
 
+## v10.0.0
+
+Changes since `v9.0.0`.
+
+This release unifies context turns under `#talk`, removes the
+deprecated `LLM::Bot` alias, and adds shared option resolution
+through `LLM::Utils`.
+
+Class-level agent tunables can now be resolved lazily via Proc,
+`Array[...]` schema/tool param types are supported, and a `key?`
+method has been added on providers.
+
+Agent tool confirmation hooks let selected tools be approved or
+cancelled before execution. Keep reading to learn more.
+
+### Breaking
+
+* **Unify context turns under `#talk`** <br>
+  Remove `LLM::Context#respond` and route responses-mode turns through
+  `LLM::Context#talk` with `mode: :responses` instead.
+
+* **Remove the `LLM::Bot` alias** <br>
+  Remove the backward-compatible `LLM::Bot` alias for `LLM::Context`.
+  Use `LLM::Context` directly instead.
+
+### Add
+
+* **Add shared option resolution through `LLM::Utils`** <br>
+  Add `LLM::Utils.resolve_option` for resolving configured values as
+  literals, procs, symbol-named methods, or duplicated hashes, and use
+  it in agent and ORM option resolution paths.
+
+* **Resolve all class-level agent tunables via Proc** <br>
+  Let `model`, `tools`, `skills`, `schema`, `stream`, and `tracer`
+  declared with a block be lazily evaluated against the agent instance
+  at initialization time, matching how `stream` and `tracer` already
+  worked.
+
+  Add `LLM::Agent#params` for direct access to the underlying context
+  parameters.
+
+  Ported from mruby-llm.
+
+* **Support `Array[...]` schema and tool param types** <br>
+  Let `LLM::Schema` properties and `LLM::Tool` params accept
+  `Array[...]` type declarations, including mixed item unions that are
+  serialized as `anyOf` array items.
+
+* **Add `LLM::Provider#key?`** <br>
+  Add `key?` to providers so callers can check whether a non-blank API
+  key has been configured.
+
+* **Add agent tool confirmation hooks** <br>
+  Add `LLM::Agent.confirm` and `LLM::Agent#on_tool_confirmation` so
+  selected tools can be approved or cancelled before execution. Pending
+  tool resolution now relies on `LLM::Context#functions` so confirmed
+  tools are not executed twice when mixed with unconfirmed tool calls.
+
+* **Add `LLM::Function#spawn(:call).wait`** <br>
+  Add task-shaped sequential execution support for direct
+  `LLM::Function#spawn(:call).wait`.
+
+### Fix
+
+* **Reduce private internal methods on `LLM::Stream`** <br>
+  Remove `tool_not_found` and `__tools__` from `LLM::Stream`. The
+  `__tools__` logic is inlined directly into `__find__` since that
+  was its only caller. The `tool_not_found` utility method was unused
+  externally and added unnecessary surface to LLM::Stream.
+
+  Ported from mruby-llm.
+
+## v9.0.0
+
+Changes since `v8.1.0`.
+
+This release deepens llm.rb's transport and cost-tracking surface. It
+replaces the old mutable `persist!` API with constructor-driven transport
+selection, removes `#call` from contexts and agents in favor of explicit
+`ctx.wait(:call)`, makes queued stream waits strategy-free, and deletes
+the unused `LLM::Utils` module.
+
+It adds cache read/write token tracking
+with corresponding cost components, audio and image token pricing,
+`LLM::Context#functions?` for queue-aware tool loops,
+`LLM::Agent.stream` DSL support, and exposes `#stream` readers on
+contexts and agents.
+
+The HTTP transport layer has been refactored around shared backends so
+providers, MCP, and custom transports all use the same normalized
+response interface.
+
+### Breaking
+
+* **Remove `#call` as a context and agent tool-loop API** <br>
+  Remove `LLM::Context#call(:functions)` and `LLM::Agent#call(:functions)`.
+  Tool loops should use `ctx.wait(:call)` or `agent.wait(:call)` instead.
+  The ActiveRecord and Sequel wrappers no longer expose `#call` passthroughs
+  for stored llm.rb contexts.
+
+* **Make HTTP transport selection constructor-driven** <br>
+  Remove public `persist!` and `.persistent` mutation APIs from
+  providers, transports, and MCP clients. Select persistent behavior at
+  construction time with `persistent: true`, `LLM::Transport.net_http`,
+  `LLM::Transport.net_http_persistent`, or an explicit `transport:`
+  override.
+
+* **Make queued stream waits strategy-free** <br>
+  Change `LLM::Stream::Queue#wait` to resolve queued work by the actual
+  task types already present in the queue instead of accepting an
+  external wait strategy. `LLM::Stream#wait(...)` remains compatible but
+  now ignores its arguments when delegating to the queue.
+
+* **Remove unused `LLM::Utils`** <br>
+  Delete the `LLM::Utils` module and remove its remaining unused
+  provider includes and top-level require.
+
+### Add
+
+* **Expose `#stream` readers on contexts and agents** <br>
+  Add public `LLM::Context#stream` and `LLM::Agent#stream` accessors so
+  callers can inspect the active stream object directly.
+
+* **Track cache read and write tokens in usage** <br>
+  Add `cache_read_tokens` and `cache_write_tokens` to `LLM::Usage` and
+  preserve them through completion usage adaptation and context usage
+  aggregation.
+
+* **Add `LLM::Context#functions?` for queue-aware tool loops** <br>
+  Add `functions?` to `LLM::Context` and the ActiveRecord and Sequel
+  wrappers so callers can detect pending tool work through either the
+  bound stream queue or unresolved functions, and update the docs to
+  prefer `while ctx.functions?` over `ctx.functions.any?` in tool-loop
+  examples.
+
+* **Add `:call` as a first-class wait strategy** <br>
+  Add `:call` to pending-function wait paths so `ctx.wait(:call)` can
+  prefer queued streamed work when present and otherwise fall back to
+  direct sequential function execution through `spawn(:call).wait`.
+
+* **Read provider cache usage into completion responses** <br>
+  Read cache read tokens from provider usage metadata, including OpenAI
+  `usage.prompt_tokens_details` and Anthropic
+  `usage.cache_read_input_tokens`. Read Anthropic cache write tokens
+  from `usage.cache_creation_input_tokens`, and expose explicit
+  zero-valued `cache_write_tokens` methods on providers that do not
+  report cache creation usage.
+
+* **Extend cost tracking with cache write pricing** <br>
+  Extend `LLM::Cost` with `cache_read_costs`, `cache_write_costs`, and
+  `reasoning_costs` alongside the existing `input_costs` and
+  `output_costs`. Add `#to_h` for structured cost insight and update
+  `ctx.cost` to calculate all available components from registry
+  pricing data.
+
+* **Price input and output audio separately** <br>
+  Track `input_audio_tokens` and `output_audio_tokens` in usage and
+  include `input_audio_costs` and `output_audio_costs` in `LLM::Cost`
+  so multimodal requests report accurate audio spend.
+
+* **Track image tokens in input cost reporting** <br>
+  Add `input_image_tokens` to usage and include `input_image_costs` in
+  `LLM::Cost` using the model's generic input rate so image-bearing
+  prompts report their input spend.
+
+* **Add `LLM::Agent.stream` DSL support** <br>
+  Let agents define a default `stream` through the class DSL, including
+  block-based stream construction so each agent instance can resolve its
+  stream the same way `tracer` does.
+
+### Change
+
+* **Refactor HTTP transports around shared backends** <br>
+  Split `Net::HTTP` and `Net::HTTP::Persistent` into separate
+  `LLM::Transport` implementations, move HTTP-specific request helpers
+  and response execution into the shared transport layer, and let MCP
+  HTTP wrap those transports instead of maintaining a separate
+  transient/persistent client split.
+
+* **Share transport overrides across providers and MCP** <br>
+  Let both provider construction and `LLM::MCP.http(...)` accept
+  `LLM::Transport` instances or classes as HTTP transport overrides, so
+  callers can reuse the same transport implementation across the
+  runtime.
+
+* **Let custom transports adapt their own response objects** <br>
+  Introduce a transport response interface so custom transports can
+  adapt backend-specific response objects to one normalized shape and
+  have them work with the existing provider execution and error-handling
+  code.
+
 ## v8.1.0
 
 Changes since `v8.0.0`.
@@ -43,7 +234,7 @@ DSML tool-marker filtering in streamed text.
   blocks that Bedrock rejects.
 
 * **Suppress Bedrock DSML tool markers in streamed text** <br>
-  Filter `"<｜DSML｜function_calls"` markers out of streamed Bedrock
+  Filter `\"<｜DSML｜function_calls\"` markers out of streamed Bedrock
   assistant text so tool-call sentinels do not leak into user-visible
   output.
 
@@ -96,8 +287,7 @@ and `acts_as_agent`.
 
 * **Allow `persistent: true` on `LLM::MCP.http`** <br>
   Let `LLM::MCP.http(...)` enable persistent HTTP transport directly
-  through `persistent: true`, instead of requiring a separate
-  `.persistent` call after construction.
+  through `persistent: true` at construction time.
 
 * **Expose `LLM::Function#runner` as public API** <br>
   Promote the internal runner instantiation to a public `runner` method on
@@ -195,7 +385,7 @@ provider usage has been recorded yet.
   buffer API.
 
 * **Support percentage compaction token thresholds** <br>
-  Let `LLM::Compactor` accept `token_threshold:` values like `"90%"` so
+  Let `LLM::Compactor` accept `token_threshold:` values like `\"90%\"` so
   compaction can trigger at a percentage of the active model context
   window.
 
@@ -978,7 +1168,7 @@ Changes since `v4.9.0`.
 
 - Add HTTP transport for MCP with `LLM::MCP::Transport::HTTP` for remote servers
 - Add JSON Schema union types (`any_of`, `all_of`, `one_of`) with parser integration
-- Add JSON Schema type array union support (e.g., `"type\": [\"object\", \"null\"]`)
+- Add JSON Schema type array union support (e.g., `\"type\": [\"object\", \"null\"]`)
 - Add JSON Schema type inference from `const`, `enum`, or `default` fields
 
 ### Change
@@ -1079,7 +1269,7 @@ Notable merged work in this range includes:
 - `Add rack + websocket example (#130)`
 - `feat(gemspec): add changelog URI (#136)`
 - `feat(function): alias ThreadGroup#wait as ThreadGroup#value (#62)`
-- README and screencast refresh across `#66`, `#67`, `#68`, `#71`, and
+- README and screencast refresh across `#66`, `#68`, `#71`, and
   `#72`
 - `chore(bot): update deprecation warning from v5.0 to v6.0`
 - `fix(deepseek): tolerate malformed tool arguments`