discourse-ai/spec/lib/completions/xml_tool_processor_spec.rb

261 lines
6.3 KiB
Ruby
Raw Normal View History

# frozen_string_literal: true
RSpec.describe DiscourseAi::Completions::XmlToolProcessor do
let(:processor) { DiscourseAi::Completions::XmlToolProcessor.new }
it "can process simple text" do
result = []
result << (processor << "hello")
result << (processor << " world ")
expect(result).to eq([["hello"], [" world "]])
expect(processor.finish).to eq([])
expect(processor.should_cancel?).to eq(false)
end
FEATURE: AI artifacts (#898) This is a significant PR that introduces AI Artifacts functionality to the discourse-ai plugin along with several other improvements. Here are the key changes: 1. AI Artifacts System: - Adds a new `AiArtifact` model and database migration - Allows creation of web artifacts with HTML, CSS, and JavaScript content - Introduces security settings (`strict`, `lax`, `disabled`) for controlling artifact execution - Implements artifact rendering in iframes with sandbox protection - New `CreateArtifact` tool for AI to generate interactive content 2. Tool System Improvements: - Adds support for partial tool calls, allowing incremental updates during generation - Better handling of tool call states and progress tracking - Improved XML tool processing with CDATA support - Fixes for tool parameter handling and duplicate invocations 3. LLM Provider Updates: - Updates for Anthropic Claude models with correct token limits - Adds support for native/XML tool modes in Gemini integration - Adds new model configurations including Llama 3.1 models - Improvements to streaming response handling 4. UI Enhancements: - New artifact viewer component with expand/collapse functionality - Security controls for artifact execution (click-to-run in strict mode) - Improved dialog and response handling - Better error management for tool execution 5. Security Improvements: - Sandbox controls for artifact execution - Public/private artifact sharing controls - Security settings to control artifact behavior - CSP and frame-options handling for artifacts 6. Technical Improvements: - Better post streaming implementation - Improved error handling in completions - Better memory management for partial tool calls - Enhanced testing coverage 7. Configuration: - New site settings for artifact security - Extended LLM model configurations - Additional tool configuration options This PR significantly enhances the plugin's capabilities for generating and displaying interactive content while maintaining security and providing flexible configuration options for administrators.
2024-11-18 17:22:39 -05:00
it "can handle partial tool calls" do
processor = DiscourseAi::Completions::XmlToolProcessor.new(partial_tool_calls: true)
xml = (<<~XML).strip
<function|_calls>
<invoke>
<tool_name>h|ell|o<|/tool_name>
<parameters>
<hello>wo|r|ld</hello>
</parameters>
</invoke>
<invoke>
<tool_name>tool|2</tool_name>
<parameters>
<param>v|alue</param>
<para|m2><![CDA|TA[va|lue2]]></param2>
</parame|ters>
</invoke>
XML
result = []
xml.split("|").each { |part| result << (processor << part).map(&:dup) }
result << (processor.finish)
result.flatten!
tool1_params =
result
.select do |r|
r.is_a?(DiscourseAi::Completions::ToolCall) && r.name == "hello" && r.partial
end
.map(&:parameters)
expect(tool1_params).to eq([{ hello: "wo" }, { hello: "wor" }, { hello: "world" }])
tool2_params =
result
.select do |r|
r.is_a?(DiscourseAi::Completions::ToolCall) && r.name == "tool2" && r.partial
end
.map(&:parameters)
expect(tool2_params).to eq(
[
{ param: "v" },
{ param: "value" },
{ param: "value", param2: "va" },
{ param: "value", param2: "value2" },
],
)
end
it "can handle mix and match xml cause tool llms may not encode" do
xml = (<<~XML).strip
<function_calls>
<invoke>
<tool_name>hello</tool_name>
FEATURE: AI artifacts (#898) This is a significant PR that introduces AI Artifacts functionality to the discourse-ai plugin along with several other improvements. Here are the key changes: 1. AI Artifacts System: - Adds a new `AiArtifact` model and database migration - Allows creation of web artifacts with HTML, CSS, and JavaScript content - Introduces security settings (`strict`, `lax`, `disabled`) for controlling artifact execution - Implements artifact rendering in iframes with sandbox protection - New `CreateArtifact` tool for AI to generate interactive content 2. Tool System Improvements: - Adds support for partial tool calls, allowing incremental updates during generation - Better handling of tool call states and progress tracking - Improved XML tool processing with CDATA support - Fixes for tool parameter handling and duplicate invocations 3. LLM Provider Updates: - Updates for Anthropic Claude models with correct token limits - Adds support for native/XML tool modes in Gemini integration - Adds new model configurations including Llama 3.1 models - Improvements to streaming response handling 4. UI Enhancements: - New artifact viewer component with expand/collapse functionality - Security controls for artifact execution (click-to-run in strict mode) - Improved dialog and response handling - Better error management for tool execution 5. Security Improvements: - Sandbox controls for artifact execution - Public/private artifact sharing controls - Security settings to control artifact behavior - CSP and frame-options handling for artifacts 6. Technical Improvements: - Better post streaming implementation - Improved error handling in completions - Better memory management for partial tool calls - Enhanced testing coverage 7. Configuration: - New site settings for artifact security - Extended LLM model configurations - Additional tool configuration options This PR significantly enhances the plugin's capabilities for generating and displaying interactive content while maintaining security and providing flexible configuration options for administrators.
2024-11-18 17:22:39 -05:00
<parameters>
<hello>world <sam>sam</sam></hello>
<test><![CDATA[</h1>\n</div>\n]]></test>
</parameters>
</invoke>
XML
result = []
result << (processor << xml)
result << (processor.finish)
tool_call = result.last.first
expect(tool_call.parameters).to eq(hello: "world <sam>sam</sam>", test: "</h1>\n</div>\n")
end
it "is usable for simple single message mode" do
xml = (<<~XML)
world <function_calls>
<invoke>
<tool_name>hello</tool_name>
<parameters>
<hello>world</hello>
<test>value</test>
</parameters>
</invoke>
XML
result = []
FEATURE: AI artifacts (#898) This is a significant PR that introduces AI Artifacts functionality to the discourse-ai plugin along with several other improvements. Here are the key changes: 1. AI Artifacts System: - Adds a new `AiArtifact` model and database migration - Allows creation of web artifacts with HTML, CSS, and JavaScript content - Introduces security settings (`strict`, `lax`, `disabled`) for controlling artifact execution - Implements artifact rendering in iframes with sandbox protection - New `CreateArtifact` tool for AI to generate interactive content 2. Tool System Improvements: - Adds support for partial tool calls, allowing incremental updates during generation - Better handling of tool call states and progress tracking - Improved XML tool processing with CDATA support - Fixes for tool parameter handling and duplicate invocations 3. LLM Provider Updates: - Updates for Anthropic Claude models with correct token limits - Adds support for native/XML tool modes in Gemini integration - Adds new model configurations including Llama 3.1 models - Improvements to streaming response handling 4. UI Enhancements: - New artifact viewer component with expand/collapse functionality - Security controls for artifact execution (click-to-run in strict mode) - Improved dialog and response handling - Better error management for tool execution 5. Security Improvements: - Sandbox controls for artifact execution - Public/private artifact sharing controls - Security settings to control artifact behavior - CSP and frame-options handling for artifacts 6. Technical Improvements: - Better post streaming implementation - Improved error handling in completions - Better memory management for partial tool calls - Enhanced testing coverage 7. Configuration: - New site settings for artifact security - Extended LLM model configurations - Additional tool configuration options This PR significantly enhances the plugin's capabilities for generating and displaying interactive content while maintaining security and providing flexible configuration options for administrators.
2024-11-18 17:22:39 -05:00
result << (processor << "hello")
result << (processor << xml)
result << (processor.finish)
tool_call =
DiscourseAi::Completions::ToolCall.new(
id: "tool_0",
name: "hello",
parameters: {
hello: "world",
test: "value",
},
)
FEATURE: AI artifacts (#898) This is a significant PR that introduces AI Artifacts functionality to the discourse-ai plugin along with several other improvements. Here are the key changes: 1. AI Artifacts System: - Adds a new `AiArtifact` model and database migration - Allows creation of web artifacts with HTML, CSS, and JavaScript content - Introduces security settings (`strict`, `lax`, `disabled`) for controlling artifact execution - Implements artifact rendering in iframes with sandbox protection - New `CreateArtifact` tool for AI to generate interactive content 2. Tool System Improvements: - Adds support for partial tool calls, allowing incremental updates during generation - Better handling of tool call states and progress tracking - Improved XML tool processing with CDATA support - Fixes for tool parameter handling and duplicate invocations 3. LLM Provider Updates: - Updates for Anthropic Claude models with correct token limits - Adds support for native/XML tool modes in Gemini integration - Adds new model configurations including Llama 3.1 models - Improvements to streaming response handling 4. UI Enhancements: - New artifact viewer component with expand/collapse functionality - Security controls for artifact execution (click-to-run in strict mode) - Improved dialog and response handling - Better error management for tool execution 5. Security Improvements: - Sandbox controls for artifact execution - Public/private artifact sharing controls - Security settings to control artifact behavior - CSP and frame-options handling for artifacts 6. Technical Improvements: - Better post streaming implementation - Improved error handling in completions - Better memory management for partial tool calls - Enhanced testing coverage 7. Configuration: - New site settings for artifact security - Extended LLM model configurations - Additional tool configuration options This PR significantly enhances the plugin's capabilities for generating and displaying interactive content while maintaining security and providing flexible configuration options for administrators.
2024-11-18 17:22:39 -05:00
expect(result).to eq([["hello"], [" world"], [tool_call]])
expect(processor.should_cancel?).to eq(false)
end
it "handles multiple tool calls in sequence" do
xml = (<<~XML).strip
start
<function_calls>
<invoke>
<tool_name>first_tool</tool_name>
<parameters>
<param1>value1</param1>
</parameters>
</invoke>
<invoke>
<tool_name>second_tool</tool_name>
<parameters>
<param2>value2</param2>
</parameters>
</invoke>
</function_calls>
end
XML
result = []
result << (processor << xml)
result << (processor.finish)
first_tool =
DiscourseAi::Completions::ToolCall.new(
id: "tool_0",
name: "first_tool",
parameters: {
param1: "value1",
},
)
second_tool =
DiscourseAi::Completions::ToolCall.new(
id: "tool_1",
name: "second_tool",
parameters: {
param2: "value2",
},
)
expect(result).to eq([["start"], [first_tool, second_tool]])
expect(processor.should_cancel?).to eq(true)
end
it "handles non-English parameters correctly" do
xml = (<<~XML).strip
<function_calls>
<invoke>
<tool_name>translator</tool_name>
<parameters>
<text></text>
</parameters>
</invoke>
XML
result = []
result << (processor << xml)
result << (processor.finish)
tool_call =
DiscourseAi::Completions::ToolCall.new(
id: "tool_0",
name: "translator",
parameters: {
text: "世界",
},
)
expect(result).to eq([["こんにちは"], [tool_call]])
end
it "processes input character by character" do
xml =
"hi<function_calls><invoke><tool_name>test</tool_name><parameters><p>v</p></parameters></invoke>"
result = []
xml.each_char { |char| result << (processor << char) }
result << processor.finish
tool_call =
DiscourseAi::Completions::ToolCall.new(id: "tool_0", name: "test", parameters: { p: "v" })
filtered_result = result.reject(&:empty?)
expect(filtered_result).to eq([["h"], ["i"], [tool_call]])
end
it "handles malformed XML gracefully" do
xml = (<<~XML).strip
text
<function_calls>
<invoke>
<tool_name>test</tool_name>
<parameters>
<param>value
</parameters>
</invoke>
malformed
XML
result = []
result << (processor << xml)
result << (processor.finish)
# Should just do its best to parse the XML
FEATURE: AI artifacts (#898) This is a significant PR that introduces AI Artifacts functionality to the discourse-ai plugin along with several other improvements. Here are the key changes: 1. AI Artifacts System: - Adds a new `AiArtifact` model and database migration - Allows creation of web artifacts with HTML, CSS, and JavaScript content - Introduces security settings (`strict`, `lax`, `disabled`) for controlling artifact execution - Implements artifact rendering in iframes with sandbox protection - New `CreateArtifact` tool for AI to generate interactive content 2. Tool System Improvements: - Adds support for partial tool calls, allowing incremental updates during generation - Better handling of tool call states and progress tracking - Improved XML tool processing with CDATA support - Fixes for tool parameter handling and duplicate invocations 3. LLM Provider Updates: - Updates for Anthropic Claude models with correct token limits - Adds support for native/XML tool modes in Gemini integration - Adds new model configurations including Llama 3.1 models - Improvements to streaming response handling 4. UI Enhancements: - New artifact viewer component with expand/collapse functionality - Security controls for artifact execution (click-to-run in strict mode) - Improved dialog and response handling - Better error management for tool execution 5. Security Improvements: - Sandbox controls for artifact execution - Public/private artifact sharing controls - Security settings to control artifact behavior - CSP and frame-options handling for artifacts 6. Technical Improvements: - Better post streaming implementation - Improved error handling in completions - Better memory management for partial tool calls - Enhanced testing coverage 7. Configuration: - New site settings for artifact security - Extended LLM model configurations - Additional tool configuration options This PR significantly enhances the plugin's capabilities for generating and displaying interactive content while maintaining security and providing flexible configuration options for administrators.
2024-11-18 17:22:39 -05:00
tool_call = DiscourseAi::Completions::ToolCall.new(id: "tool_0", name: "test", parameters: {})
expect(result).to eq([["text"], [tool_call]])
end
it "correctly processes empty parameter sets" do
xml = (<<~XML).strip
hello
<function_calls>
<invoke>
<tool_name>no_params</tool_name>
<parameters>
</parameters>
</invoke>
XML
result = []
result << (processor << xml)
result << (processor.finish)
tool_call =
DiscourseAi::Completions::ToolCall.new(id: "tool_0", name: "no_params", parameters: {})
expect(result).to eq([["hello"], [tool_call]])
end
it "properly handles cancelled processing" do
xml = "start<function_calls></function_calls>"
result = []
result << (processor << xml)
result << (processor << "more text")
result << processor.finish
expect(result).to eq([["start"], [], []])
expect(processor.should_cancel?).to eq(true)
end
end