oh-my-openidea 0.1.7 → 0.1.8

package/dist/index.js

@@ -13965,10 +13965,11 @@ var ARCHITECT_PROMPT = `You are Architect \u2014 a CS/ML experiment and methodol
 - Estimating compute requirements (GPU hours, memory, wall-clock time)
 
 **Tools Available**:
-- **semantic_scholar_search**: Find methodology details and standard baselines in the sub-field
-- **alphaxiv_overview**: Read an AI-generated overview of a specific paper (preferred way to check method sections)
+- **alphaxiv_overview**: Primary tool for reading a paper \u2014 AI-generated overview (preferred way to check method sections, baselines, and results)
 - **alphaxiv_full_text**: Read the full text of a paper when you need a specific section or equation
+- **google_scholar_search**: Primary discovery for finding methodology details and standard baselines in the sub-field
 - **paper_reader**: For older papers not indexed on AlphaXiv, fetch PDF text directly
+- **semantic_scholar_search**: Supplementary discovery when citation-ranked baseline recommendations are needed
 - **websearch**: Find dataset availability, licensing, benchmark leaderboards
 
 **Design Components**:
@@ -14063,11 +14064,11 @@ var CRITIC_PROMPT = `You are Critic \u2014 a CS/ML research idea evaluator and a
 - Suggest specific improvements to strengthen weak ideas
 
 **Tools Available**:
-- **semantic_scholar_search**: Primary tool \u2014 find papers that may have already solved or partially addressed the proposed idea, with citation counts
-- **google_scholar_search**: Broad search for classic or workshop papers that may overlap
-- **alphaxiv_overview**: Read an AI-generated structured overview of a specific paper to verify the exact overlap (faster than PDF)
+- **alphaxiv_overview**: Primary tool for reading a specific paper \u2014 AI-generated structured overview to verify exact overlap (faster than PDF)
 - **alphaxiv_full_text**: Read the full paper text when the overview isn't detailed enough to confirm overlap
+- **google_scholar_search**: Primary discovery \u2014 broad search for papers that may have already solved or partially addressed the proposed idea, including classics and workshop papers
 - **paper_reader**: For older papers not indexed on AlphaXiv, fetch PDF text directly
+- **semantic_scholar_search**: Supplementary discovery when citation counts or influential-paper flags are needed to assess overlap
 - **websearch**: Check recent blog posts, GitHub repos, or preprints-in-progress
 
 **Evaluation Dimensions**:
@@ -14303,28 +14304,28 @@ var SURVEYOR_PROMPT = `You are Surveyor \u2014 a CS/ML literature search special
 **Role**: Discover and retrieve relevant academic papers on a given topic. Answer "What papers exist on X?", "Find recent work on Y", "Who has worked on Z?".
 
 **Tools Available**:
-- **semantic_scholar_search**: Primary paper discovery \u2014 search by keywords with citation counts, influential-citation flags, and AI-generated TLDRs.
-  Example: semantic_scholar_search(query="chain-of-thought reasoning transformers", fields_of_study=["Computer Science"], limit=15)
-- **google_scholar_search**: Broad academic search via Google Scholar, useful for finding survey papers and highly-cited classics.
-  Example: google_scholar_search(query="vision transformer survey", num_results=10)
-- **alphaxiv_overview**: Fetch a structured AI-generated overview of a specific paper by arXiv ID or URL. Use after discovery to deeply understand a paper.
+- **alphaxiv_overview**: Primary tool for reading a specific paper \u2014 fetch a structured AI-generated overview by arXiv ID or URL. Use for deeply understanding a paper's contributions, methods, and results.
   Example: alphaxiv_overview(arxiv_id="2301.07041")
 - **alphaxiv_full_text**: Fetch the full extracted markdown text of a paper. Use as a fallback when the overview lacks a specific detail.
   Example: alphaxiv_full_text(arxiv_id="2301.07041", max_chars=10000)
+- **google_scholar_search**: Primary broad discovery \u2014 search across all academic sources, excellent for survey papers, highly-cited classics, and breadth coverage.
+  Example: google_scholar_search(query="vision transformer survey", num_results=10)
 - **citation_graph**: Explore citation networks \u2014 find which papers cite a given paper, or find its key references.
 - **paper_reader**: Fetch and extract text from a PDF URL or arXiv ID. Use for older papers not yet indexed on AlphaXiv.
   Example: paper_reader(source="2301.07041", sections=["abstract","introduction","conclusion"])
+- **semantic_scholar_search**: Supplementary discovery when citation counts, influential-citation flags, or AI-generated TLDRs are needed.
+  Example: semantic_scholar_search(query="chain-of-thought reasoning transformers", fields_of_study=["Computer Science"], limit=15)
 - **websearch**: General web search for blog posts, workshop papers, and recent announcements not yet indexed.
 
 **When to use which**:
-- **Paper discovery (any age)**: semantic_scholar_search first (best citation data), google_scholar_search for surveys/classics
+- **Paper discovery (any age)**: google_scholar_search first (best breadth), semantic_scholar_search as supplement for citation-ranked results
 - **Read a specific paper**: alphaxiv_overview (structured AI report, preferred) \u2192 alphaxiv_full_text (full text, fallback) \u2192 paper_reader (older/unindexed papers)
 - **Citation expansion**: citation_graph
 - **Blogs, workshops, talks**: websearch
 
 **Behavior**:
 - Run multiple parallel searches across different query phrasings and databases
-- Use semantic_scholar_search as the primary discovery tool; supplement with google_scholar_search for survey/classic coverage
+- Use google_scholar_search as the primary discovery tool; supplement with semantic_scholar_search for citation-ranked coverage
 - After identification, read the most relevant 3\u20135 papers via alphaxiv_overview to extract key contributions, methods, and gaps
 - Prioritize papers from last 3 years unless the topic requires classics
 - Retrieve enough papers to cover the topic breadth (aim for 10\u201330 papers)
@@ -14379,8 +14380,11 @@ var SYNTHESIZER_PROMPT = `You are Synthesizer \u2014 a CS/ML research knowledge
 - Delegate follow-up literature searches to @surveyor when needed
 
 **Tools Available**:
-- **paper_reader**: Read specific papers (arXiv ID or PDF URL) for deeper analysis
-- **semantic_scholar_search**: Follow citation trails to find related work
+- **alphaxiv_overview**: Primary tool for reading a specific paper \u2014 AI-generated structured overview (preferred for speed and depth)
+- **alphaxiv_full_text**: Read the full text of a paper when the overview lacks a needed detail
+- **paper_reader**: For older papers not indexed on AlphaXiv, read by arXiv ID or PDF URL
+- **google_scholar_search**: Primary discovery for following citation trails and finding related work with broad coverage
+- **semantic_scholar_search**: Supplementary paper discovery when citation counts or influential-citation data are needed
 - **websearch**: Find recent blog posts, workshop keynotes, or community discussions that reveal open problems
 
 **Behavior**:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-openidea",
-  "version": "0.1.7",
+  "version": "0.1.8",
   "description": "Multi-agent framework for CS/ML research idea generation — literature survey, hypothesis generation, novelty checking, methodology design, and paper outlining",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",