| license: apache-2.0 | |
| language: en | |
| tags: | |
| - agentic-ai | |
| - planning | |
| - task-decomposition | |
| - dag | |
| - resource-estimation | |
| pipeline_tag: any-to-any | |
| # AIF-Planner | |
| **Task Decomposition & Dependency Resolution Engine** — Converts high-level objectives into executable directed acyclic graphs with resource estimation and constraint validation. | |
| ## Capabilities | |
| - **Objective Parsing**: Natural language to structured task graph | |
| - **Dependency Resolution**: Automatic ordering with cycle detection | |
| - **Resource Estimation**: Token count, time, and cost projections per task | |
| - **Constraint Validation**: Policy and capability checks before dispatch | |
| - **Plan Optimization**: Merge parallelizable tasks; eliminate redundant steps | |
| - **Alternative Generation**: Produce multiple execution strategies with cost/benefit analysis | |
| ## Planning Pipeline | |
| ``` | |
| Objective → Decompose → Resolve Dependencies → Estimate Resources → Validate → Optimize → DAG | |
| ``` | |
| ## Integration | |
| The Planner feeds directly into the AIF-Orchestrator for execution. Plans are versioned and stored for audit replay. | |
| ```python | |
| from runtime.enterprise_control_plane import Planner | |
| planner = Planner() | |
| plan = planner.decompose("Build and deploy a customer-facing chatbot with Stripe integration") | |
| # Returns: DAG with 12 tasks, 4 parallel branches, estimated 850 tokens | |
| ``` | |
| ## Repository | |
| [https://github.com/frostyjay7813/AIF](https://github.com/frostyjay7813/AIF) | |