@apmantza/greedysearch-pi 1.4.2 → 1.5.0

package/newfeaturesideas.md

@@ -1,105 +0,0 @@
-# New Feature Ideas
-
-Ideas for future features — thinking from the perspective of an AI assistant using these tools.
-
----
-
-## 1. Source Verification
-
-**Problem:** I get sources but can't verify if they're live, updated, or actually support the claimed content.
-
-```
-verify_sources({ urls: ["https://...", "https://..."] })
-→ [{ url, status: 200, title, snippet, lastModified, claim: "supports X" }]
-```
-
-**Use cases:**
-- Before citing a source, verify it's not 404
-- Check if a page actually contains the claimed information
-- Get last-modified dates to assess freshness
-
----
-
-## 2. Incremental / Continuation Research
-
-**Problem:** After deep_research on "RAG vs fine-tuning", going deeper on just RAG means re-running everything with a new query and losing original context.
-
-```
-deep_research({ query: "RAG vs fine-tuning", ... })  // initial
-continue_research({ previousId: "...", query: "production RAG architectures" })  // goes deeper on RAG
-```
-
-**Use cases:**
-- Drill into a specific aspect after initial broad research
-- Build on previous results without re-fetching everything
-- Progressive disclosure of complex topics
-
----
-
-## 3. Multi-Query Synthesis
-
-**Problem:** One query isn't enough for complex research. I chain multiple greedy_search calls manually.
-
-```
-multi_research({
-  queries: ["auth best practices", "NextAuth vs Clerk vs Lucia", "Next.js auth security"],
-  synthesize: true
-})
-```
-
-**Use cases:**
-- "Best auth for Next.js" needs multiple angles
-- Research with different facets (comparison, security, performance)
-- Casting a wider net when single query returns narrow results
-
----
-
-## 4. Structured Extraction
-
-**Problem:** When researching "which libraries are maintained", I want tables (name, stars, last commit, license), not prose.
-
-```
-extract_structured({
-  query: "Python HTTP client libraries 2026",
-  schema: { name: "string", stars: "number", lastUpdated: "date", async: "boolean" }
-})
-```
-
-**Use cases:**
-- Library comparisons as structured data
-- Dependency audits
-- Feature matrices for tools/frameworks
-
----
-
-## 5. Confidence Scoring on Specific Claims
-
-**Problem:** I say "high confidence" but it's hand-wavy. What if I could ask: "how confident are we that library X is actively maintained?"
-
-```
-verify_claim({
-  claim: "Prisma is actively maintained",
-  evidence: ["last commit: 2 weeks ago", "open issues: 45", "npm downloads: 2M/week"]
-})
-→ { confidence: 0.95, reasoning: "..." }
-```
-
----
-
-## 6. Research Cache / History
-
-**Problem:** I do expensive deep_research, then the user asks a follow-up. I have to re-run everything.
-
-```
-get_research(id: "...")  // retrieve previous results
-list_research({ query: "RAG" })  // find related previous research
-```
-
----
-
-## Priority
-
-1. **Source verification** — high value, relatively simple, fixes trust gap
-2. **Multi-query synthesis** — high value, complex but powerful
-3. **Incremental research** — medium value, nice UX improvement
-4. **Structured extraction** — medium value, specialized use cases