rcrewai 0.10.0 → 0.11.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (56) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGELOG.md +24 -0
  3. data/README.md +133 -23
  4. data/ROADMAP.md +1 -1
  5. data/docs/api/agent.md +5 -2
  6. data/docs/api/configuration.md +36 -41
  7. data/docs/api/crew.md +34 -48
  8. data/docs/api/index.md +7 -0
  9. data/docs/api/task.md +10 -91
  10. data/docs/api/tools.md +19 -0
  11. data/docs/examples/api-integration.md +4 -4
  12. data/docs/examples/async-execution.md +4 -4
  13. data/docs/examples/code-review-crew.md +1 -1
  14. data/docs/examples/content-marketing-pipeline.md +3 -3
  15. data/docs/examples/custom-tools.md +8 -8
  16. data/docs/examples/customer-support.md +5 -5
  17. data/docs/examples/data-analysis-team.md +4 -4
  18. data/docs/examples/database-operations.md +4 -4
  19. data/docs/examples/ecommerce-operations.md +4 -4
  20. data/docs/examples/financial-analysis.md +3 -3
  21. data/docs/examples/hierarchical-crew.md +3 -2
  22. data/docs/examples/production-ready-crew.md +16 -10
  23. data/docs/examples/research-development.md +5 -5
  24. data/docs/examples/simple-research-crew.md +2 -2
  25. data/docs/examples/social-media.md +4 -4
  26. data/docs/examples/task-automation.md +2 -2
  27. data/docs/examples/tool-composition.md +42 -27
  28. data/docs/examples/web-scraping.md +6 -6
  29. data/docs/index.md +61 -44
  30. data/docs/tutorials/advanced-agents.md +194 -917
  31. data/docs/tutorials/agent-options.md +1 -1
  32. data/docs/tutorials/checkpointing.md +65 -0
  33. data/docs/tutorials/concurrency.md +57 -0
  34. data/docs/tutorials/custom-tools.md +198 -1143
  35. data/docs/tutorials/deployment.md +156 -155
  36. data/docs/tutorials/getting-started.md +69 -81
  37. data/docs/tutorials/index.md +19 -0
  38. data/docs/tutorials/interceptors-and-events.md +84 -0
  39. data/docs/tutorials/live-streaming.md +52 -0
  40. data/docs/tutorials/multiple-crews.md +70 -51
  41. data/docs/tutorials/providers.md +115 -0
  42. data/docs/upgrading-to-0.11.md +55 -0
  43. data/lib/rcrewai/agent.rb +0 -2
  44. data/lib/rcrewai/configuration.rb +64 -29
  45. data/lib/rcrewai/context_window.rb +12 -2
  46. data/lib/rcrewai/llm_clients/anthropic.rb +60 -9
  47. data/lib/rcrewai/llm_clients/base.rb +7 -0
  48. data/lib/rcrewai/llm_clients/bedrock.rb +29 -2
  49. data/lib/rcrewai/llm_clients/google.rb +36 -4
  50. data/lib/rcrewai/llm_clients/ollama.rb +17 -3
  51. data/lib/rcrewai/llm_clients/openai.rb +14 -1
  52. data/lib/rcrewai/llm_clients/openai_responses.rb +21 -2
  53. data/lib/rcrewai/pricing.rb +13 -3
  54. data/lib/rcrewai/tool_runner.rb +5 -1
  55. data/lib/rcrewai/version.rb +1 -1
  56. metadata +7 -1
checksums.yaml CHANGED
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  SHA256:
3
- metadata.gz: 1e33f9998ef9f32aa53d4c6df3e8e7b04314f7b5cbb9fe06e5c7d1936b466222
4
- data.tar.gz: 294a6a4a5ed7d5f1b5fbb58c3edd0bb69a0cd5d9e61da07b9f8dabf6c4457414
3
+ metadata.gz: 70f424175fd6d8116947f503da80a59b3de84f12b9f3c308361e1104b1ebba06
4
+ data.tar.gz: cd6e8d8f22d6617efb0d9ccee5c2a7064c3aa6c92403c35edb717e26358369b1
5
5
  SHA512:
6
- metadata.gz: f64195cd3065e2c26dd65ddea08246bec905754192e8ba0fad1a23f7bcbfc998c2f8a4d3cad91f7491f5d8e0f2fc5c4cd53be60de47e737f9d41b2c898089ba7
7
- data.tar.gz: 5107c4953f3ff80f86125fdbb5325bf582b1c624b07e97b739890fd21d0c83c892c7adff4fd982556141dd5a60b186814bcc3b1644056a89d589788810fee860
6
+ metadata.gz: f0ac2ac824ed422e7d9918aae99e7d811ee115bc30e19e5fe1d54c53e0a77d9af0229d296e483f638fa9f3a376c1304bbe6933c7bae233c54b3f8101a5e9a665
7
+ data.tar.gz: 9b3bb92c5af6e0d5258ae34daf48454d87dbcbc062e7dbc578aa764d4a6765a4023c72eb4e7f01d2ffa71920a743cf88115b7f484a2432f7f6cde5a07b922dcf
data/CHANGELOG.md CHANGED
@@ -7,6 +7,30 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
7
7
 
8
8
  ## [Unreleased]
9
9
 
10
+ ## [0.11.0] - 2026-11-05
11
+
12
+ Makes agents with tools work on real providers, and current Claude models usable. Requests after a tool call were malformed on every provider since 0.3.0, because specs stubbed the LLM and never checked the real request.
13
+
14
+ **Not yet proven against live APIs.** Request formats are covered by specs against stubbed HTTP. If a provider rejects a tool loop, please [open an issue](https://github.com/gkosmo/rcrewAI/issues).
15
+
16
+ Contains behavior changes. See `docs/upgrading-to-0.11.md`.
17
+
18
+ ### Fixed
19
+ - **Behavior change:** `config.model = '...'` was silently ignored for `:openai`, `:anthropic`, `:google` and `:azure`, because each provider's default model took precedence over it. A model now resolves as an explicit `<provider>_model`, then `model`, then the provider default. If your config sets `model` and you have been unknowingly running on a default, you will now get the model you asked for.
20
+ - A global `config.model` no longer follows an agent that switches provider with `llm: :anthropic`; that agent uses its own provider's model.
21
+ - `:ollama` no longer requires an API key in `RCrewAI.configure`, so a local setup works without a placeholder key. A model is still required.
22
+ - `:openai_responses` ignored `openai_model` and always sent `gpt-4`, and its `api_key` ignored `openai_api_key`. It now shares both with `:openai`.
23
+ - `Configuration#supported_providers` listed 5 providers; it now reports all 9.
24
+ - **Tool loops sent malformed history on every provider, since 0.3.0.** Specs stubbed the LLM, so the real requests were never checked. OpenAI, Azure, OpenAI-compatible and Snowflake got `tool_calls` without the `type` and `function` wrapper. The Responses API, Ollama, Google and Bedrock dropped the calls or sent results as text. Each now sends its provider's own format, and `raw_content` replays Gemini thought signatures and Responses and Bedrock reasoning unchanged.
25
+ - **Anthropic tool loops sent malformed history.** After a tool call the client dropped `tool_calls`, sent the assistant turn as `content: null`, and sent results as plain user text. It now sends `tool_use` and `tool_result` blocks, with one turn's results in one user message. Responses carry `raw_content`, which `ToolRunner` replays so thinking blocks come back unchanged.
26
+ - **Current Claude models rejected every request,** because the client always sent `temperature` and the default was `0.1`. It now sends only a temperature you set. `Configuration#explicit_temperature` reports it. Two internal calls that hardcoded `temperature: 0.2` no longer do, so on other providers they use your configured temperature.
27
+ - `with_overrides(temperature: nil)` and `llm: { temperature: nil }` now clear a global temperature; `nil` used to be ignored.
28
+ - Corrected Anthropic prices that were out of date (`claude-opus-4-7`, `claude-haiku-4-5`) and Anthropic context windows.
29
+
30
+ ### Changed
31
+ - **Behavior change:** `:anthropic` now defaults to `claude-opus-5-5` and `:google` to `gemini-3.8-flash`, replacing retired models. Added prices and context windows for current Claude and Gemini models.
32
+ - **Behavior change:** providers with no default model (`:ollama`, `:openai_compatible`, `:bedrock`, `:snowflake`) now raise `ConfigurationError` when no model is set, instead of sending `gpt-4`, which those endpoints would not generally serve.
33
+
10
34
  ## [0.10.0] - 2026-10-19
11
35
 
12
36
  ### Added
data/README.md CHANGED
@@ -11,7 +11,7 @@ RCrewAI is a Ruby implementation of the CrewAI framework, allowing you to create
11
11
  ## 🚀 Features
12
12
 
13
13
  - **🤖 Intelligent Agents**: AI agents with reasoning loops, memory, and tool usage capabilities
14
- - **🔗 Multi-LLM Support**: OpenAI, Anthropic (Claude), Google (Gemini), Azure OpenAI, and Ollama
14
+ - **🔗 Multi-LLM Support**: OpenAI (Chat Completions and Responses), Anthropic (Claude), Google (Gemini), Azure OpenAI, Ollama, AWS Bedrock, Snowflake Cortex, and any OpenAI-compatible endpoint
15
15
  - **🛠️ Rich Tool Ecosystem**: Web search, file operations, SQL, email, code execution, PDF processing, and custom tools
16
16
  - **🧠 Cognitive Memory**: Semantic recall (embeddings + cosine) with optional SQLite persistence and short-term/long-term/entity/tool memory types
17
17
  - **🤝 Human-in-the-Loop**: Interactive approval workflows, human guidance, and collaborative decision making
@@ -21,6 +21,10 @@ RCrewAI is a Ruby implementation of the CrewAI framework, allowing you to create
21
21
  - **🎯 Flexible Orchestration**: Sequential, hierarchical, and concurrent execution modes
22
22
  - **🌊 Flows**: Event-driven workflows with `start`/`listen`/`router`, branching, and persistent state
23
23
  - **📚 Knowledge (RAG)**: Ground agents in your own documents with built-in retrieval
24
+ - **💾 Checkpointing**: Resume a failed run without repeating finished tasks
25
+ - **🔍 Observability**: LLM request and response hooks, plus a nested event stream
26
+ - **🚦 Bounded Concurrency**: Parallel tool calls, consensus rounds and embeddings, each with a limit
27
+ - **🎙️ Live Streaming**: Cancellable Anthropic streams with connection reuse
24
28
  - **💎 Ruby-First Design**: Built specifically for Ruby developers with idiomatic patterns
25
29
 
26
30
  ## 📦 Installation
@@ -134,43 +138,70 @@ puts "✅ Crew completed #{results[:completed_tasks]}/#{results[:total_tasks]} t
134
138
  ## 🔧 LLM Provider Support
135
139
 
136
140
  ```ruby
137
- # OpenAI (GPT-4, GPT-3.5, etc.)
141
+ # OpenAI
138
142
  RCrewAI.configure do |config|
139
143
  config.llm_provider = :openai
140
144
  config.openai_api_key = ENV['OPENAI_API_KEY']
141
- config.model = 'gpt-4'
145
+ config.openai_model = 'gpt-4o'
142
146
  end
143
147
 
144
- # Anthropic Claude
148
+ # Anthropic
145
149
  RCrewAI.configure do |config|
146
150
  config.llm_provider = :anthropic
147
- config.anthropic_api_key = ENV['ANTHROPIC_API_KEY']
148
- config.model = 'claude-3-sonnet-20240229'
151
+ config.anthropic_api_key = ENV['ANTHROPIC_API_KEY']
152
+ config.anthropic_model = 'claude-opus-5-5'
149
153
  end
150
154
 
151
155
  # Google Gemini
152
156
  RCrewAI.configure do |config|
153
157
  config.llm_provider = :google
154
158
  config.google_api_key = ENV['GOOGLE_API_KEY']
155
- config.model = 'gemini-pro'
159
+ config.google_model = 'gemini-3.8-flash'
156
160
  end
157
161
 
158
162
  # Azure OpenAI
159
163
  RCrewAI.configure do |config|
160
164
  config.llm_provider = :azure
161
165
  config.azure_api_key = ENV['AZURE_OPENAI_API_KEY']
162
- config.azure_endpoint = ENV['AZURE_OPENAI_ENDPOINT']
163
- config.model = 'gpt-4'
166
+ config.base_url = ENV['AZURE_OPENAI_ENDPOINT']
167
+ config.deployment_name = 'your-deployment'
168
+ config.api_version = '2024-10-21'
164
169
  end
165
170
 
166
- # Local Ollama
171
+ # Local Ollama (no API key)
167
172
  RCrewAI.configure do |config|
168
173
  config.llm_provider = :ollama
169
- config.ollama_url = 'http://localhost:11434'
170
- config.model = 'llama2'
174
+ config.base_url = 'http://localhost:11434'
175
+ config.model = 'llama3.1'
176
+ end
177
+
178
+ # Any OpenAI-compatible endpoint: Together, Groq, Fireworks, vLLM, LiteLLM, OpenRouter
179
+ RCrewAI.configure do |config|
180
+ config.llm_provider = :openai_compatible
181
+ config.base_url = 'https://api.groq.com/openai/v1'
182
+ config.api_key = ENV['GROQ_API_KEY']
183
+ config.model = 'your-model'
184
+ end
185
+
186
+ # AWS Bedrock (bearer token; SigV4 signing isn't built in)
187
+ RCrewAI.configure do |config|
188
+ config.llm_provider = :bedrock
189
+ config.aws_region = 'us-east-1'
190
+ config.api_key = ENV['BEDROCK_API_KEY']
191
+ config.model = 'your-bedrock-model-id'
192
+ end
193
+
194
+ # Snowflake Cortex
195
+ RCrewAI.configure do |config|
196
+ config.llm_provider = :snowflake
197
+ config.snowflake_account = 'your-account'
198
+ config.api_key = ENV['SNOWFLAKE_TOKEN']
199
+ config.model = 'your-cortex-model'
171
200
  end
172
201
  ```
173
202
 
203
+ There's also `:openai_responses`, which uses OpenAI's Responses API (no streaming) with your OpenAI model and key settings.
204
+
174
205
  ### Per-agent LLM
175
206
 
176
207
  The `RCrewAI.configure` block sets the crew-wide default. Any agent can override
@@ -184,7 +215,7 @@ researcher = RCrewAI::Agent.new(name: 'researcher', role: '...', goal: '...',
184
215
 
185
216
  # Provider + model (and optionally api_key / temperature)
186
217
  manager = RCrewAI::Agent.new(name: 'manager', role: '...', goal: '...',
187
- llm: { provider: :anthropic, model: 'claude-3-opus-20240229' })
218
+ llm: { provider: :anthropic, model: 'claude-opus-5-5' })
188
219
 
189
220
  worker = RCrewAI::Agent.new(name: 'worker', role: '...', goal: '...',
190
221
  llm: { provider: :openai, model: 'gpt-4o-mini' })
@@ -240,7 +271,7 @@ the executing agent a head start:
240
271
  crew = RCrewAI::Crew.new('research_crew', planning: true)
241
272
  # Optionally use a dedicated (e.g. stronger) planner model:
242
273
  crew = RCrewAI::Crew.new('research_crew', planning: true,
243
- planning_llm: { provider: :anthropic, model: 'claude-3-opus-20240229' })
274
+ planning_llm: { provider: :anthropic, model: 'claude-opus-5-5' })
244
275
  ```
245
276
 
246
277
  Planning is best-effort: if the planner errors or returns unparseable output,
@@ -465,6 +496,73 @@ becomes the result (ties break toward the task's assigned agent). A proposer tha
465
496
  errors is dropped; if all proposals fail the task is marked failed. Consensus
466
497
  multiplies LLM calls, so `consensus_agents` bounds the cost on larger crews.
467
498
 
499
+ ## 💾 Checkpointing
500
+
501
+ Save progress after each task, then resume a failed run without paying for finished work again:
502
+
503
+ ```ruby
504
+ store = RCrewAI::Checkpoint::FileStore.new('.rcrewai/checkpoints')
505
+
506
+ crew.execute(checkpoint: store)
507
+ run_id = crew.run_id
508
+
509
+ # Later, on a crew with the same task names:
510
+ crew.resume(run_id, checkpoint: store)
511
+ RCrewAI::Checkpoint.lineage(store, crew.run_id) # => [original_id, resumed_id]
512
+ ```
513
+
514
+ See the [checkpointing tutorial](docs/tutorials/checkpointing.md).
515
+
516
+ ## 🔍 Interceptors & Events
517
+
518
+ Hook into every LLM call, or subscribe to what the crew is doing:
519
+
520
+ ```ruby
521
+ client = RCrewAI::LLMClient.for_provider(:openai, RCrewAI.configuration,
522
+ before_request: ->(payload, _ctx) { payload.merge(user: 'anonymous') })
523
+ client.after_response { |_result, ctx| puts "#{ctx[:model]}: #{ctx[:duration_ms]}ms"; nil }
524
+
525
+ agent = RCrewAI::Agent.new(name: 'analyst', role: 'Analyst', goal: 'Analyze', llm: client)
526
+
527
+ crew.execute do |event|
528
+ print event.text if event.is_a?(RCrewAI::Events::TextDelta)
529
+ end
530
+ ```
531
+
532
+ Events carry `id` and `parent_id`, so you can rebuild a run into a trace. See [interceptors and events](docs/tutorials/interceptors-and-events.md).
533
+
534
+ ## 🚦 Concurrency
535
+
536
+ Tool calls in one turn, consensus rounds, and embeddings on `:google` and `:ollama` run concurrently, each limited to 8 at a time by default.
537
+
538
+ ```ruby
539
+ # For an agent whose tools aren't thread-safe, or that calls a rate-limited API:
540
+ RCrewAI::Agent.new(name: 'clerk', role: 'Clerk', goal: 'Update records', parallel_tools: false)
541
+
542
+ RCrewAI::Crew.new('panel', process: :consensual, consensus_max_concurrency: 4)
543
+ RCrewAI::Knowledge::Embedder.new(provider: :ollama, max_concurrency: 2)
544
+ ```
545
+
546
+ See the [concurrency tutorial](docs/tutorials/concurrency.md).
547
+
548
+ ## 🎙️ Live Streaming
549
+
550
+ For real-time callers such as voice agents, the Anthropic client can cancel a stream mid-reply. It also reuses its connection between turns:
551
+
552
+ ```ruby
553
+ client = RCrewAI::LLMClient.for_provider(:anthropic)
554
+ token = RCrewAI::CancelToken.new
555
+
556
+ begin
557
+ client.chat(messages: messages, model: 'claude-haiku-5-5', cancel: token,
558
+ stream: ->(event) { print event.text if event.is_a?(RCrewAI::Events::TextDelta) })
559
+ rescue RCrewAI::StreamCancelled
560
+ puts '(interrupted)'
561
+ end
562
+ ```
563
+
564
+ See the [live streaming tutorial](docs/tutorials/live-streaming.md).
565
+
468
566
  ## 💡 Examples
469
567
 
470
568
  ### Hierarchical Team with Human Oversight
@@ -530,31 +628,39 @@ results = crew.execute(
530
628
  ## 🛠️ CLI Usage
531
629
 
532
630
  ```bash
533
- # Create a new crew
534
- $ rcrewai new my_research_crew --process sequential
631
+ # Create and list crews
632
+ $ rcrewai new my_research_crew
633
+ $ rcrewai list
535
634
 
536
- # Create agents with tools
635
+ # Create agents and tasks
537
636
  $ rcrewai agent new researcher \
538
637
  --role "Senior Research Analyst" \
539
- --tools web_search,file_writer \
540
- --human-input
638
+ --goal "Find reliable sources"
541
639
 
542
- # Create tasks with dependencies
543
640
  $ rcrewai task new research \
544
641
  --description "Research latest AI developments" \
545
642
  --agent researcher \
546
643
  --async
547
644
 
548
- # Run crews
549
- $ rcrewai run --crew my_research_crew --async
645
+ # Run a crew
646
+ $ rcrewai run --crew my_research_crew
647
+
648
+ # Inspect saved checkpoints
649
+ $ rcrewai checkpoint list
650
+ $ rcrewai checkpoint info RUN_ID
651
+ $ rcrewai checkpoint delete RUN_ID
550
652
  ```
551
653
 
654
+ `checkpoint` and `version` are fully working. The `new`, `list`, `run`, `agent` and `task` commands are placeholders: they don't save anything, the lists are sample names, and `run` executes an empty crew. For now, define crews in Ruby.
655
+
552
656
  ## 📚 Examples & Documentation
553
657
 
554
658
  - **[Getting Started Guide](docs/tutorials/getting-started.md)**: Learn the basics
555
659
  - **[Human-in-the-Loop Example](examples/human_in_the_loop_example.rb)**: Interactive AI workflows
556
660
  - **[Hierarchical Teams](examples/hierarchical_crew_example.rb)**: Manager coordination
557
661
  - **[Async Execution](examples/async_execution_example.rb)**: Performance optimization
662
+ - **[LLM Providers](docs/tutorials/providers.md)**: Setup for all nine providers
663
+ - **[Upgrading to 0.11](docs/upgrading-to-0.11.md)**: What to check when upgrading
558
664
  - **[API Documentation](docs/api/)**: Complete API reference
559
665
 
560
666
  ## 🎯 Use Cases
@@ -616,10 +722,14 @@ gem 'rcrewai-rails'
616
722
  Rails.application.routes.draw do
617
723
  mount RcrewAI::Rails::Engine, at: '/rcrewai'
618
724
  end
725
+ ```
619
726
 
727
+ ```bash
620
728
  # Generate a new crew
621
- rails generate rcrew_ai:crew marketing_crew
729
+ $ rails generate rcrewai:rails:crew marketing_crew
730
+ ```
622
731
 
732
+ ```ruby
623
733
  # Create persistent agents and tasks through Rails models
624
734
  crew = RcrewAI::Rails::Crew.create!(name: "Content Team", description: "AI content generation")
625
735
  agent = crew.agents.create!(name: "writer", role: "Content Writer", goal: "Create engaging content")
data/ROADMAP.md CHANGED
@@ -5,7 +5,7 @@ This roadmap tracks feature parity between **RCrewAI** (Ruby) and the upstream
5
5
 
6
6
  ## Current status
7
7
 
8
- - **RCrewAI:** `0.8.1` released; 1.0.0 concurrency work merged to `main`, unreleased
8
+ - **RCrewAI:** `0.11.0` released
9
9
  - **Upstream crewai:** `1.15.21`
10
10
 
11
11
  RCrewAI is a faithful port of CrewAI's **"Crews"** mental model (Agents / Tasks /
data/docs/api/agent.md CHANGED
@@ -24,10 +24,11 @@ Creates a new agent instance.
24
24
  - `allow_delegation` (Boolean, optional) - Allow agent to delegate tasks (default: false)
25
25
  - `manager` (Boolean, optional) - Designate as manager agent (default: false)
26
26
  - `max_iterations` (Integer, optional) - Maximum reasoning iterations (default: 10)
27
- - `max_execution_time` (Integer, optional) - Maximum execution time in seconds (default: 300)
27
+ - `max_execution_time` (Integer, optional) - Stored but not currently enforced (default: 300)
28
28
  - `human_input` (Boolean, optional) - Enable human-in-the-loop interactions (default: false)
29
29
  - `require_approval_for_tools` (Boolean, optional) - Require human approval for tool usage
30
30
  - `require_approval_for_final_answer` (Boolean, optional) - Require human approval for final results
31
+ - `parallel_tools` (Boolean, optional) - Run several tool calls from one turn concurrently (default: true). See [Concurrency]({{ site.baseurl }}/tutorials/concurrency)
31
32
  - `llm` (Symbol \| Hash \| client, optional) - Per-agent LLM override: a provider symbol (`:anthropic`), an options hash (`{ provider:, model:, api_key:, temperature: }`), or a pre-built client. Defaults to the global configuration. See [Advanced Agent Options]({{ site.baseurl }}/tutorials/agent-options)
32
33
  - `reasoning` (Boolean, optional) - Run a reasoning/planning pass before answering (default: false); surfaced on the result as `:reasoning`
33
34
  - `max_reasoning_attempts` (Integer, optional) - Reasoning retries on empty output (default: 3)
@@ -55,7 +56,7 @@ agent = RCrewAI::Agent.new(
55
56
 
56
57
  ## Instance Methods
57
58
 
58
- ### `#execute_task(task)`
59
+ ### `#execute_task(task, stream: nil, **opts)`
59
60
 
60
61
  Executes a given task using the agent's reasoning capabilities.
61
62
 
@@ -258,6 +259,8 @@ Maximum reasoning iterations allowed.
258
259
  **Type:** Integer
259
260
 
260
261
  ### `max_execution_time`
262
+
263
+ **Not currently enforced.** The value is stored, but nothing stops an agent when it's exceeded.
261
264
  Maximum execution time in seconds.
262
265
 
263
266
  **Type:** Integer
@@ -12,11 +12,16 @@ The Configuration class manages settings for various LLM providers and global op
12
12
 
13
13
  RCrewAI supports the following LLM providers:
14
14
 
15
- - **OpenAI** (GPT-4, GPT-3.5-turbo, etc.)
16
- - **Anthropic** (Claude 3, Claude 2, etc.)
17
- - **Google** (Gemini Pro, PaLM, etc.)
18
- - **Azure OpenAI** (Azure-hosted OpenAI models)
19
- - **Ollama** (Local/self-hosted models)
15
+ - **OpenAI** (`:openai`, and `:openai_responses` for the Responses API)
16
+ - **Anthropic** (`:anthropic`)
17
+ - **Google Gemini** (`:google`)
18
+ - **Azure OpenAI** (`:azure`)
19
+ - **Ollama** (`:ollama`, local models, no API key)
20
+ - **OpenAI-compatible endpoints** (`:openai_compatible`)
21
+ - **AWS Bedrock** (`:bedrock`)
22
+ - **Snowflake Cortex** (`:snowflake`)
23
+
24
+ See [LLM Providers]({{ site.baseurl }}/tutorials/providers) for setup examples.
20
25
 
21
26
  ## Basic Configuration
22
27
 
@@ -45,7 +50,7 @@ end
45
50
  RCrewAI.configure do |config|
46
51
  config.llm_provider = :openai
47
52
  config.openai_api_key = "your-openai-key"
48
- config.openai_model = "gpt-4"
53
+ config.openai_model = "gpt-4o"
49
54
  config.temperature = 0.1
50
55
  config.max_tokens = 4000
51
56
  end
@@ -53,60 +58,43 @@ end
53
58
 
54
59
  ## Provider-Specific Configuration
55
60
 
61
+ Model names change often, so check your provider's documentation for what's available.
62
+
56
63
  ### OpenAI
57
64
 
58
65
  ```ruby
59
66
  RCrewAI.configure do |config|
60
67
  config.llm_provider = :openai
61
68
  config.openai_api_key = ENV['OPENAI_API_KEY']
62
- config.openai_model = 'gpt-4' # or 'gpt-3.5-turbo', 'gpt-4-turbo'
69
+ config.openai_model = 'gpt-4o'
63
70
  config.temperature = 0.1
64
71
  config.max_tokens = 4000
65
72
  end
66
73
  ```
67
74
 
68
- **Available Models:**
69
- - `gpt-4`
70
- - `gpt-4-turbo`
71
- - `gpt-3.5-turbo`
72
- - `gpt-3.5-turbo-16k`
73
-
74
75
  ### Anthropic (Claude)
75
76
 
76
77
  ```ruby
77
78
  RCrewAI.configure do |config|
78
79
  config.llm_provider = :anthropic
79
80
  config.anthropic_api_key = ENV['ANTHROPIC_API_KEY']
80
- config.anthropic_model = 'claude-3-sonnet-20240229'
81
- config.temperature = 0.1
81
+ config.anthropic_model = 'claude-opus-5-5'
82
82
  config.max_tokens = 4000
83
83
  end
84
84
  ```
85
85
 
86
- **Available Models:**
87
- - `claude-3-opus-20240229`
88
- - `claude-3-sonnet-20240229`
89
- - `claude-3-haiku-20240307`
90
- - `claude-2.1`
91
- - `claude-2.0`
92
-
93
86
  ### Google (Gemini)
94
87
 
95
88
  ```ruby
96
89
  RCrewAI.configure do |config|
97
90
  config.llm_provider = :google
98
91
  config.google_api_key = ENV['GOOGLE_API_KEY']
99
- config.google_model = 'gemini-pro'
92
+ config.google_model = 'gemini-3.8-flash'
100
93
  config.temperature = 0.1
101
94
  config.max_tokens = 4000
102
95
  end
103
96
  ```
104
97
 
105
- **Available Models:**
106
- - `gemini-pro`
107
- - `gemini-pro-vision`
108
- - `text-bison-001`
109
-
110
98
  ### Azure OpenAI
111
99
 
112
100
  ```ruby
@@ -114,7 +102,7 @@ RCrewAI.configure do |config|
114
102
  config.llm_provider = :azure
115
103
  config.azure_api_key = ENV['AZURE_OPENAI_API_KEY']
116
104
  config.base_url = "https://your-resource.openai.azure.com/"
117
- config.api_version = "2023-05-15"
105
+ config.api_version = "2024-10-21"
118
106
  config.deployment_name = "your-deployment-name"
119
107
  config.azure_model = 'gpt-4'
120
108
  end
@@ -170,9 +158,15 @@ RCrewAI.reset_configuration!
170
158
  | Option | Type | Default | Description |
171
159
  |--------|------|---------|-------------|
172
160
  | `llm_provider` | Symbol | `:openai` | LLM provider to use |
173
- | `temperature` | Float | `0.1` | Sampling temperature (0.0 - 2.0) |
161
+ | `temperature` | Float | `0.1` | Sampling temperature (0.0 - 2.0). Anthropic only receives a value you set |
174
162
  | `max_tokens` | Integer | `4000` | Maximum tokens in response |
175
163
  | `timeout` | Integer | `120` | Request timeout in seconds |
164
+ | `base_url` | String | none | Endpoint for `:azure`, `:openai_compatible`, `:ollama` and `:anthropic` gateways (`LLM_BASE_URL`) |
165
+ | `aws_region` | String | none | Region for `:bedrock` (`AWS_REGION`, `AWS_DEFAULT_REGION`) |
166
+ | `snowflake_account` | String | none | Account for `:snowflake` (`SNOWFLAKE_ACCOUNT`) |
167
+ | `pricing` | Hash | none | Per-model price overrides |
168
+ | `ollama_native_tools` | Boolean | none | Override native tool detection for Ollama models |
169
+ | `log_level` | Symbol | `:info` | Log level |
176
170
 
177
171
  ### Provider-Specific Keys
178
172
 
@@ -182,15 +176,19 @@ RCrewAI.reset_configuration!
182
176
  | `anthropic_api_key` | `ANTHROPIC_API_KEY` | Anthropic API key |
183
177
  | `google_api_key` | `GOOGLE_API_KEY` | Google API key |
184
178
  | `azure_api_key` | `AZURE_OPENAI_API_KEY` | Azure OpenAI API key |
179
+ | `api_key` | `LLM_API_KEY` | Fallback key, and the key for `:openai_compatible`, `:bedrock` and `:snowflake` |
185
180
 
186
181
  ### Provider-Specific Models
187
182
 
188
183
  | Option | Default | Description |
189
184
  |--------|---------|-------------|
190
185
  | `openai_model` | `gpt-4` | OpenAI model name |
191
- | `anthropic_model` | `claude-3-sonnet-20240229` | Anthropic model name |
192
- | `google_model` | `gemini-pro` | Google model name |
186
+ | `anthropic_model` | `claude-opus-5-5` | Anthropic model name |
187
+ | `google_model` | `gemini-3.8-flash` | Google model name |
193
188
  | `azure_model` | `gpt-4` | Azure OpenAI model name |
189
+ | `model` | none | Used when the provider-specific option isn't set, and required for `:ollama`, `:openai_compatible`, `:bedrock` and `:snowflake` |
190
+
191
+ The `gpt-4` defaults are old, so set a model explicitly for OpenAI and Azure. A model resolves as the provider-specific option, then `model`, then the default.
194
192
 
195
193
  ### Azure-Specific Settings
196
194
 
@@ -236,7 +234,7 @@ export GOOGLE_API_KEY="AIza..."
236
234
  export AZURE_OPENAI_API_KEY="..."
237
235
 
238
236
  # Azure Specific
239
- export AZURE_API_VERSION="2023-05-15"
237
+ export AZURE_API_VERSION="2024-10-21"
240
238
  export AZURE_DEPLOYMENT_NAME="gpt-4-deployment"
241
239
  export LLM_BASE_URL="https://your-resource.openai.azure.com/"
242
240
 
@@ -251,7 +249,7 @@ RCrewAI validates configuration when `configure` is called:
251
249
  ```ruby
252
250
  begin
253
251
  RCrewAI.configure do |config|
254
- config.llm_provider = :invalid_provider
252
+ config.llm_provider = nil
255
253
  end
256
254
  rescue RCrewAI::ConfigurationError => e
257
255
  puts "Configuration error: #{e.message}"
@@ -283,12 +281,9 @@ RCrewAI.reset_configuration!
283
281
 
284
282
  1. **Use Environment Variables**: Store API keys in environment variables, not in code
285
283
  2. **Provider Selection**: Choose providers based on task requirements:
286
- - OpenAI GPT-4: General-purpose, high quality
287
- - Claude: Great for analysis and reasoning
288
- - Gemini: Good for creative tasks
289
- - Azure: Enterprise requirements
284
+ - Azure, Bedrock, Snowflake: Your organization already uses that platform
290
285
  - Ollama: Privacy-sensitive or offline use
291
- 3. **Temperature Settings**:
286
+ 3. **Temperature Settings** (not accepted by current Claude models):
292
287
  - Low (0.1-0.3): Factual, consistent outputs
293
288
  - Medium (0.5-0.7): Balanced creativity and consistency
294
289
  - High (0.8-1.0): Creative, varied outputs
@@ -312,7 +307,7 @@ end
312
307
  ```ruby
313
308
  # Check supported providers
314
309
  puts RCrewAI.configuration.supported_providers
315
- # => [:openai, :anthropic, :google, :azure, :ollama]
310
+ # => [:openai, :anthropic, :google, :azure, :ollama, :openai_compatible, :bedrock, :snowflake, :openai_responses]
316
311
  ```
317
312
 
318
313
  **Azure Configuration Issues**
@@ -321,7 +316,7 @@ RCrewAI.configure do |config|
321
316
  config.llm_provider = :azure
322
317
  config.azure_api_key = ENV['AZURE_OPENAI_API_KEY']
323
318
  config.base_url = "https://your-resource.openai.azure.com/" # Include trailing slash
324
- config.api_version = "2023-05-15" # Use supported API version
319
+ config.api_version = "2024-10-21" # Use supported API version
325
320
  config.deployment_name = "your-deployment-name" # Exact deployment name
326
321
  end
327
322
  ```