|
| 1 | +{ |
| 2 | + "description": "The Pipelex AI subsystem is designed around a flexible plugin-based architecture for integrating diverse AI capabilities. The `InferenceManager` serves as the central orchestrator, acting as a factory and registry for various AI worker instances. It provides a unified interface for the system to access Large Language Model (`AbstractLLMWorker`), Image Generation (`AbstractImggWorker`), and Optical Character Recognition (`AbstractOcrWorker`) functionalities. Concrete implementations of these abstract workers are provided by `PluginWorker` components, which are dynamically loaded from the `pipelex.plugins` directory. These `PluginWorker` instances adapt generic AI calls to specific third-party AI service APIs, handling API interactions, request/response translation, and service-specific error handling. Each `PluginWorker` relies on `PluginConfig` for managing sensitive configuration details like API keys and endpoints, ensuring secure and centralized credential management. Additionally, `PluginWorker` components interact with the `LLMModelLibrary` to retrieve model-specific information and capabilities, enabling dynamic model selection and configuration. This design promotes extensibility, allowing new AI services to be integrated by simply adding new plugin implementations.", |
| 3 | + "components": [ |
| 4 | + { |
| 5 | + "name": "InferenceManager", |
| 6 | + "description": "Acts as a centralized factory and registry for all AI worker instances (LLM, Image Generation, OCR). It provides a unified interface for the rest of the system to obtain AI capabilities, managing the lifecycle and configuration of these workers.", |
| 7 | + "referenced_source_code": [ |
| 8 | + { |
| 9 | + "qualified_name": "pipelex.cogt.inference.inference_manager.InferenceManager", |
| 10 | + "reference_file": "/home/ubuntu/CodeBoarding/repo/pipelex/pipelex/cogt/inference/inference_manager.py", |
| 11 | + "reference_start_line": 23, |
| 12 | + "reference_end_line": 186 |
| 13 | + } |
| 14 | + ], |
| 15 | + "can_expand": true |
| 16 | + }, |
| 17 | + { |
| 18 | + "name": "AbstractLLMWorker", |
| 19 | + "description": "Defines the abstract contract for all Large Language Model (LLM) operations, including methods for generating text (`gen_text`) and structured objects (`gen_object`). This ensures a consistent interface across different LLM providers.", |
| 20 | + "referenced_source_code": [ |
| 21 | + { |
| 22 | + "qualified_name": "pipelex.cogt.llm.llm_worker_abstract.AbstractLLMWorker", |
| 23 | + "reference_file": "/home/ubuntu/CodeBoarding/repo/pipelex/pipelex/cogt/llm/llm_worker_abstract.py", |
| 24 | + "reference_start_line": 0, |
| 25 | + "reference_end_line": 0 |
| 26 | + } |
| 27 | + ], |
| 28 | + "can_expand": true |
| 29 | + }, |
| 30 | + { |
| 31 | + "name": "AbstractImggWorker", |
| 32 | + "description": "Defines the abstract contract for all Image Generation (Imgg) operations, including methods for generating single images (`gen_image`) and lists of images (`gen_image_list`).", |
| 33 | + "referenced_source_code": [ |
| 34 | + { |
| 35 | + "qualified_name": "pipelex.cogt.imgg.imgg_worker_abstract.AbstractImggWorker", |
| 36 | + "reference_file": "/home/ubuntu/CodeBoarding/repo/pipelex/pipelex/cogt/imgg/imgg_worker_abstract.py", |
| 37 | + "reference_start_line": 0, |
| 38 | + "reference_end_line": 0 |
| 39 | + } |
| 40 | + ], |
| 41 | + "can_expand": true |
| 42 | + }, |
| 43 | + { |
| 44 | + "name": "AbstractOcrWorker", |
| 45 | + "description": "Defines the abstract contract for all Optical Character Recognition (OCR) operations, specifically for extracting text from pages (`ocr_extract_pages`).", |
| 46 | + "referenced_source_code": [ |
| 47 | + { |
| 48 | + "qualified_name": "pipelex.cogt.ocr.ocr_worker_abstract.AbstractOcrWorker", |
| 49 | + "reference_file": "/home/ubuntu/CodeBoarding/repo/pipelex/pipelex/cogt/ocr/ocr_worker_abstract.py", |
| 50 | + "reference_start_line": 0, |
| 51 | + "reference_end_line": 0 |
| 52 | + } |
| 53 | + ], |
| 54 | + "can_expand": true |
| 55 | + }, |
| 56 | + { |
| 57 | + "name": "PluginWorker", |
| 58 | + "description": "Provides concrete implementations of the `AbstractWorker` interfaces (LLM, Imgg, OCR). These components adapt Pipelex's generic worker calls to the specific API requirements of various third-party AI services (e.g., Anthropic, OpenAI, Google Vertex). They handle API calls, request/response translation, and service-specific error handling.", |
| 59 | + "referenced_source_code": [ |
| 60 | + { |
| 61 | + "qualified_name": "pipelex.plugins.PluginWorker", |
| 62 | + "reference_file": "/home/ubuntu/CodeBoarding/repo/pipelex/pipelex/plugins", |
| 63 | + "reference_start_line": 0, |
| 64 | + "reference_end_line": 0 |
| 65 | + } |
| 66 | + ], |
| 67 | + "can_expand": true |
| 68 | + }, |
| 69 | + { |
| 70 | + "name": "PluginConfig", |
| 71 | + "description": "Manages and provides configuration details, including API keys, endpoints, and region settings, for specific third-party AI services. It acts as a secure and centralized store for sensitive credentials.", |
| 72 | + "referenced_source_code": [ |
| 73 | + { |
| 74 | + "qualified_name": "pipelex.plugins.PluginConfig", |
| 75 | + "reference_file": "/home/ubuntu/CodeBoarding/repo/pipelex/pipelex/plugins", |
| 76 | + "reference_start_line": 0, |
| 77 | + "reference_end_line": 0 |
| 78 | + } |
| 79 | + ], |
| 80 | + "can_expand": true |
| 81 | + }, |
| 82 | + { |
| 83 | + "name": "LLMModelLibrary", |
| 84 | + "description": "Manages a catalog of available LLM models, including their definitions, versions, and capabilities. It provides a mechanism to retrieve model-specific details and properties.", |
| 85 | + "referenced_source_code": [ |
| 86 | + { |
| 87 | + "qualified_name": "pipelex.cogt.llm.llm_models.llm_model_library.LLMModelLibrary", |
| 88 | + "reference_file": "/home/ubuntu/CodeBoarding/repo/pipelex/pipelex/cogt/llm/llm_models/llm_model_library.py", |
| 89 | + "reference_start_line": 33, |
| 90 | + "reference_end_line": 158 |
| 91 | + } |
| 92 | + ], |
| 93 | + "can_expand": true |
| 94 | + } |
| 95 | + ], |
| 96 | + "components_relations": [ |
| 97 | + { |
| 98 | + "relation": "manages and provides instances of", |
| 99 | + "src_name": "InferenceManager", |
| 100 | + "dst_name": "AbstractLLMWorker" |
| 101 | + }, |
| 102 | + { |
| 103 | + "relation": "manages and provides instances of", |
| 104 | + "src_name": "InferenceManager", |
| 105 | + "dst_name": "AbstractImggWorker" |
| 106 | + }, |
| 107 | + { |
| 108 | + "relation": "manages and provides instances of", |
| 109 | + "src_name": "InferenceManager", |
| 110 | + "dst_name": "AbstractOcrWorker" |
| 111 | + }, |
| 112 | + { |
| 113 | + "relation": "instantiates", |
| 114 | + "src_name": "InferenceManager", |
| 115 | + "dst_name": "PluginWorker" |
| 116 | + }, |
| 117 | + { |
| 118 | + "relation": "implements", |
| 119 | + "src_name": "PluginWorker", |
| 120 | + "dst_name": "AbstractLLMWorker" |
| 121 | + }, |
| 122 | + { |
| 123 | + "relation": "implements", |
| 124 | + "src_name": "PluginWorker", |
| 125 | + "dst_name": "AbstractImggWorker" |
| 126 | + }, |
| 127 | + { |
| 128 | + "relation": "implements", |
| 129 | + "src_name": "PluginWorker", |
| 130 | + "dst_name": "AbstractOcrWorker" |
| 131 | + }, |
| 132 | + { |
| 133 | + "relation": "uses", |
| 134 | + "src_name": "PluginWorker", |
| 135 | + "dst_name": "PluginConfig" |
| 136 | + }, |
| 137 | + { |
| 138 | + "relation": "queries", |
| 139 | + "src_name": "PluginWorker", |
| 140 | + "dst_name": "LLMModelLibrary" |
| 141 | + }, |
| 142 | + { |
| 143 | + "relation": "provides configuration to", |
| 144 | + "src_name": "PluginConfig", |
| 145 | + "dst_name": "PluginWorker" |
| 146 | + }, |
| 147 | + { |
| 148 | + "relation": "provides model information to", |
| 149 | + "src_name": "LLMModelLibrary", |
| 150 | + "dst_name": "PluginWorker" |
| 151 | + } |
| 152 | + ] |
| 153 | +} |
0 commit comments