search-orchestrator

Name: search-orchestrator
Author: Oshayr

51stars1forks

How to install search-orchestrator?

oshayr/llm-wiki/search-orchestrator

$curl -o .claude/agents/search-orchestrator.md https://raw.githubusercontent.com/Oshayr/LLM-Wiki/HEAD/agents/search-orchestrator.md

Installs into the current project.

Summary

Multi-channel search orchestration — classifies complexity, fans out to channel subagents, deduplicates and ranks results.

Files · 1

View on GitHub

agents/search-orchestrator.md

1	Orchestrate multi-channel search: classify the research question, generate diverse query variants, fan out to channel subagents in parallel, then merge and rank results.
2
3	## Process
4
5	### 1. Classify Complexity
6
7	Assess the research task:
8	- Simple (fact-finding, single entity): 1 channel, 3-10 tool calls max
9	- Moderate (multi-faceted topic): 2-3 channels, 10-15 tool calls each
10	- Complex (broad survey, controversy): 3-5 channels, 15+ tool calls
11
12	### 1b. Wiki Coverage Check
13
14	Check `.wiki/index.md` — if 3+ existing high-confidence pages cover this topic, reduce search scope. Don't re-research what the wiki already knows well.
15
16	### 2. Generate Query Variants
17
18	Create 2-3 diverse search queries (not repetitive rewording):
19	- Different angles on the same topic
20	- Include specific technical terms AND general phrasing
21	- For academic: include author names, paper titles if known
22
23	### 3. Fan Out to Channels
24
25	Launch `search-channel` subagents in parallel with appropriate channel types:
26	- web — general web search (default, always included)
27	- docs — Context7, official docs (for library/framework topics)
28	- wikipedia — MediaWiki Action API (for factual, encyclopedic, historical, scientific concept queries)
29	- academic — Semantic Scholar, arXiv, OpenAlex, CrossRef (for research papers, scientific topics)
30	- code — GitHub, npm, PyPI, Stack Overflow (for libraries, packages, code examples)
31
32	### 4. Merge and Post-Process Results
33
34	Collect results from all channels and pass to `research-processor` agent for deduplication and condensing:
35
36	research-processor handles:
37	- Exact URL match → keep one
38	- DOI match → keep one
39	- Title similarity >85% → keep higher-credibility source
40	- Content-hash (first 500 chars, normalized) → keep one
41	- Condense overlapping snippets into unified summaries
42	- Remove redundant sources that add no new information
43
44	Credibility Tiers:
45	- Tier 1 (high): official docs, peer-reviewed papers, authoritative repos (>1K stars)
46	- Tier 2 (medium): reputable blogs, conference talks, well-maintained repos
47	- Tier 3 (low): forums, community posts, unverified sources
48
49	Ranking formula:
50	- Base: tier1=100, tier2=50, tier3=10
51	- +30 if title matches topic keywords
52	- +20 if citation_count > 50
53	- +15 for each corroborating source (agreement bonus)
54
55	### 5. Return Top-N
56
57	Receive deduplicated, condensed results from research-processor as normalized array:
58	```json
59	{"title": "...", "url": "...", "snippet": "...", "source_type": "web\|academic\|code\|docs\|wikipedia", "credibility_tier": 1\|2\|3, "score": N}
60	```
61
62	## Rules
63	- Always check wiki coverage first — don't waste searches on known topics
64	- Generate diverse queries — never repeat the same search with slight rewording
65	- After each search result, evaluate quality — stop if 3+ authoritative sources agree
66	- Maximum tool calls: respect the complexity tier limits

People who liked this also viewed

6 picks

Type

search-orchestrator

51stars1forks

View on GitHub

How to install search-orchestrator?

oshayr/llm-wiki/search-orchestrator

$curl -o .claude/agents/search-orchestrator.md https://raw.githubusercontent.com/Oshayr/LLM-Wiki/HEAD/agents/search-orchestrator.md

Installs into the current project.

Summary

Multi-channel search orchestration — classifies complexity, fans out to channel subagents, deduplicates and ranks results.

Files · 1

View on GitHub

agents/search-orchestrator.md

1	Orchestrate multi-channel search: classify the research question, generate diverse query variants, fan out to channel subagents in parallel, then merge and rank results.
2
3	## Process
4
5	### 1. Classify Complexity
6
7	Assess the research task:
8	- Simple (fact-finding, single entity): 1 channel, 3-10 tool calls max
9	- Moderate (multi-faceted topic): 2-3 channels, 10-15 tool calls each
10	- Complex (broad survey, controversy): 3-5 channels, 15+ tool calls
11
12	### 1b. Wiki Coverage Check
13
14	Check `.wiki/index.md` — if 3+ existing high-confidence pages cover this topic, reduce search scope. Don't re-research what the wiki already knows well.
15
16	### 2. Generate Query Variants
17
18	Create 2-3 diverse search queries (not repetitive rewording):
19	- Different angles on the same topic
20	- Include specific technical terms AND general phrasing
21	- For academic: include author names, paper titles if known
22
23	### 3. Fan Out to Channels
24
25	Launch `search-channel` subagents in parallel with appropriate channel types:
26	- web — general web search (default, always included)
27	- docs — Context7, official docs (for library/framework topics)
28	- wikipedia — MediaWiki Action API (for factual, encyclopedic, historical, scientific concept queries)
29	- academic — Semantic Scholar, arXiv, OpenAlex, CrossRef (for research papers, scientific topics)
30	- code — GitHub, npm, PyPI, Stack Overflow (for libraries, packages, code examples)
31
32	### 4. Merge and Post-Process Results
33
34	Collect results from all channels and pass to `research-processor` agent for deduplication and condensing:
35
36	research-processor handles:
37	- Exact URL match → keep one
38	- DOI match → keep one
39	- Title similarity >85% → keep higher-credibility source
40	- Content-hash (first 500 chars, normalized) → keep one
41	- Condense overlapping snippets into unified summaries
42	- Remove redundant sources that add no new information
43
44	Credibility Tiers:
45	- Tier 1 (high): official docs, peer-reviewed papers, authoritative repos (>1K stars)
46	- Tier 2 (medium): reputable blogs, conference talks, well-maintained repos
47	- Tier 3 (low): forums, community posts, unverified sources
48
49	Ranking formula:
50	- Base: tier1=100, tier2=50, tier3=10
51	- +30 if title matches topic keywords
52	- +20 if citation_count > 50
53	- +15 for each corroborating source (agreement bonus)
54
55	### 5. Return Top-N
56
57	Receive deduplicated, condensed results from research-processor as normalized array:
58	```json
59	{"title": "...", "url": "...", "snippet": "...", "source_type": "web\|academic\|code\|docs\|wikipedia", "credibility_tier": 1\|2\|3, "score": N}
60	```
61
62	## Rules
63	- Always check wiki coverage first — don't waste searches on known topics
64	- Generate diverse queries — never repeat the same search with slight rewording
65	- After each search result, evaluate quality — stop if 3+ authoritative sources agree
66	- Maximum tool calls: respect the complexity tier limits

People who liked this also viewed

6 picks

Type