Use Cases
across different integration architectures. Each represents a distinct pattern — from large-scale
federated search infrastructure to domain-scoped conversational agents to evidence-synthesis
pipelines — and each is running in production on the same API surface documented in this guide.
1. GGKP AI Search — Federated Search at Scale
pilot that demonstrates the federated index operating at real-world content volume. The portal
currently indexes over 200,000 knowledge products drawn from across the GGKP partner network
— spanning the Green Finance, Green Industry, and Green Policy Platforms, GEF ISLANDS, GEF FARM,
SPAR6C, and several additional partner repositories — and exposes them through a single query
interface.
receive an AI-generated summary response alongside ranked document results. The platform is
explicitly positioned as a pilot: it validates search quality and semantic relevance at scale
before those capabilities are embedded in individual partner portals.
| Endpoint | Role |
|---|---|
POST /search-federator/search | Federated keyword search across all indexed sources |
POST /search-federator/semantic | AI-powered semantic search with intent understanding |
POST /generic/api/generic | AI-generated summary response above search results |
Keyword Search
Semantic Search Across the Full Corpus
becomes practically significant at 200,000+ products, where relevant documents that do not contain
the exact query terms would be invisible to keyword search alone.
2. GEF ISLANDS & GEF FARM — Semantic Search with a Domain-Scoped Librarian Bot
GEF FARM: https://www.geffarm.org/gef-farm-knowledge
adapted to the content scope and user community of each programme. The pattern combines two API
capabilities: semantic search over a programme-specific corpus, and an embedded Knowledge
Assistant — a local librarian bot — whose responses are grounded exclusively in that programme's
indexed knowledge.
developing states. Programme officers and country-level stakeholders use the GEF Learn section to
navigate a specialist corpus where the relevant vocabulary differs meaningfully from general green
economy terminology. The librarian bot allows users to ask questions in plain language and receive
synthesised, cited answers without needing to understand the underlying taxonomy or document
structure.
management across GEF-supported countries. Content entering the portal is automatically classified
on ingestion via
/classify, with taxonomy scope constrained to the FARM domain, so that searchrelevance and assistant grounding remain tightly aligned with the programme's thematic boundaries.
Automatic Classifier, Taxonomy Management, Product Management.
| Endpoint | Role |
|---|---|
POST /search-federator/semantic | Programme-scoped semantic search |
POST /generic/api/generic | Domain-scoped librarian bot (system-prompted) |
POST /classify | Automated classification on content ingestion |
GET /taxonomy-manage/view-domainname-data | Domain-filtered vocabulary for ISLANDS or FARM |
POST /search-federator/upload-products | Indexing classified products into the federated corpus |
Semantic Search Scoped to GEF ISLANDS
Domain-Scoped Knowledge Assistant (Librarian Bot)
system parameter is the mechanism that scopes the assistant to its domain. Without it,the assistant draws on the full GGKP knowledge base; with it, responses are constrained to the
programme corpus and carry the appropriate editorial framing for the platform's users.
Automated Classification on Content Ingestion
lightweight review workflow before products are committed to the federated index:
{
"status": "success",
"data": {
"sectors": [
{
"name": "Agriculture",
"score": 0.91,
"reasoning": "Document addresses smallholder farming practices in island agricultural systems"
}
],
"themes": [
{
"name": "Biodiversity",
"score": 0.94,
"reasoning": "Central focus on ecosystem preservation and species protection strategies"
}
],
"regions": [
{
"name": "Asia-Pacific",
"score": 0.89,
"reasoning": "Case studies drawn from Pacific island nations"
}
]
},
"meta": {
"config": {
"active_taxonomy": ["Sectors", "Themes", "Regions", "Knowledge"],
"threshold": 0.80,
"max_results": 5
},
"total_time": 2.31
}
}3. Waste Management Policy Support System — Capacity Assessment and Knowledge Recommendations
takes countries from structured self-assessment through gap identification to targeted knowledge
recommendations. Each stage is distinct in its function, and the third depends directly on the
outputs of the first two — making this a genuinely sequential pipeline rather than a set of
independently usable tools.
questionnaire that captures their current waste management capacity across six components. The
questionnaire is self-administered, meaning that national counterparts report their own situation
directly into the system without requiring external facilitation. The data collected is
country-specific, sensitive in some contexts, and stored securely — access is scoped to the
submitting country and authorised platform administrators.
the platform analyses the responses to identify where capacity falls below expected thresholds
across the six components. This produces a structured gap profile: a machine-readable
representation of what needs to improve, at what severity, and in which thematic areas. This
profile is the direct input to Stage 3.
queries the GGKP Knowledge Assistant to retrieve a curated set of knowledge products from across
the GGKP network that are relevant to addressing that specific capacity deficit. The assistant
grounds its output in the indexed knowledge base, ensuring that every recommendation is
traceable to a specific document, and surfaces resources that country teams may not have
encountered through general search.
/generic/api/generic), Search Federator(
/search-federator/semantic), Taxonomy (Themes — Waste Management, Circular Economy).| Endpoint | Role in Pipeline |
|---|---|
POST /search-federator/semantic | Retrieval of domain-relevant knowledge products by gap theme |
POST /generic/api/generic | Gap-specific knowledge recommendation generation |
Selecting the Right Integration Pattern
different user needs and content contexts.
| Pattern | When to Use | Core Endpoints |
|---|---|---|
| Federated search portal | Large multi-source corpus; users need cross-boundary discovery | POST /search-federator/search, POST /search-federator/semantic |
| Domain-scoped librarian bot | Specialist programme with defined knowledge boundaries; users need guidance, not just results | POST /search-federator/semantic, POST /generic/api/generic (system-scoped) |
| Evidence-based recommendation pipeline | Policy or advisory context; recommendations must be traceable to verified sources | POST /search-federator/semantic, POST /search-federator/search, POST /generic/api/generic (RAG pattern) |
hybrid architectures are straightforward to construct. Partners whose integration requirements
do not map cleanly onto one of these patterns are encouraged to contact the GGKP technical team.