Enhance eSim Copilot with Semantic Routing, Robust RAG, and OCR-Based Value Extraction

[HARIOM-BEEP]

Related Issues

Purpose

This pull request introduces eSim Copilot, an AI-powered conversational assistant integrated into eSim to help users analyze circuits, understand schematics, and navigate eSim workflows interactively.

Prior to this change, eSim did not provide an intelligent assistant for answering user questions, interpreting schematic images, or explaining simulation inputs such as netlists. Users were required to manually inspect files or search documentation.

This PR adds a context-aware chatbot designed specifically for electronics design and simulation tasks within eSim.

---

Features Added

1. Documentation-Grounded Question Answering (RAG)

• Indexed official eSim documentation using semantic embeddings.
• Questions related to eSim are answered using relevant documentation sections.
• Prevents hallucinated responses by grounding answers in reference material.
• Falls back to a general language model when documentation is not applicable.

2. Semantic Conversation Understanding

• Maintains short conversational context to support follow-up questions.
• Automatically detects semantic topic changes using embedding similarity.
• Resets context when the user switches topics, ensuring accurate responses to unrelated questions.

3. Image-Based Circuit Analysis

• Supports analysis of schematic images provided by the user.
• Identifies circuit topology, components, and structural patterns.
• Integrates optical character recognition (OCR) to extract component labels and values from images.
• Parsed OCR-derived values are reused for follow-up questions, ensuring accurate numeric reporting.

4. Netlist Analysis and Explanation

• Supports analysis of SPICE/Ngspice netlists provided by the user.
• Identifies components, node connections, and circuit structure from raw netlists.
• Explains the function of the circuit and highlights potential issues.
• Enables users to ask follow-up questions about specific netlist elements.

5. Robust Vision, OCR, and Netlist Integration

• OCR output is injected into vision model prompts to improve recognition accuracy.
• Numeric component values are extracted directly from OCR results rather than inferred.
• Netlist content is analyzed deterministically before being passed to the language model.
• Ensures consistent and reliable analysis across images and text-based inputs.

6. Modular and Extensible Design

• Clear separation between documentation retrieval, vision analysis, OCR extraction, netlist parsing, and language model reasoning.
• Defensive fallbacks to handle incomplete or ambiguous input gracefully.
• Architecture supports future extensions such as cross-validation between schematic images and netlists.

---

Result

With this change, eSim gains a fully integrated AI assistant that:

• Answers eSim-related questions using official documentation when available
• Analyzes schematic images and accurately extracts component values
• Parses and explains netlists in a user-friendly manner
• Handles follow-up questions and topic changes correctly
• Provides general explanations when documentation or structured input is not relevant

This significantly improves usability for new users while remaining helpful for advanced users.

---

Notes

• This feature is optional and does not affect existing simulation or schematic workflows.
• No changes were made to core simulation engines or project file formats.