← Index

Wireframe In Progress

Notebook

Round 1 Created 2026-05-23 Section Analyze · Notebook

Desktop

Egan prod

Search graph, accounts… ⌘K

◆Brief

◇Accounts

TargetsIntelligenceDiscoveryPlans

◇Knowledge

GraphTaxonomyComplianceOntologies

◇Analyze

Notebook Reports Benchmarks Similarity Assistant Query

◇Operate

PipelinesSourcesIntegrationsQualityAdmin

G. Felicebuild 4a7c92

Analyze›Notebook Acme PMP — supply chain analysis

nb-204 · Python 3.13 · Piggie 0.8 · 12 cells · 8 code · 4 md · Last run 8 min ago

Created G. Felice · 2026-05-21

[1] code Python · ran 8m ago · 0.42s

# Pull Acme's neighborhood from the graph, depth 2 from piggie import connect from egan import entities, taxonomy p = connect("prod") acme = entities.get("acme-corp") nbrs = p.cypher(""" MATCH (org:Organization {id: $id})-[r*1..2]->(n) RETURN org, r, n """, id=acme.id).to_df()

out DataFrame · 247 rows × 6 cols · 12 ms

idx	org_id	rel	node_id	node_type	depth
0	acme-corp	depends_on	supplier-tier-1	Organization	1
1	acme-corp	depends_on	logistics-partner	Organization	1
2	acme-corp	depends_on	component-mfg	Organization	1
3	acme-corp	enables	distribution-net	Capability	1
4	supplier-tier-1	supplies	raw-material-x	Material	2
…	…	…	…	…	…

247 rows · memory 14.2 KB · partition prod.graph.manufacturing

[2] md Markdown · edited 14m ago

Approach

Pull the 2-hop neighborhood from Acme, then run pgvector similarity on the supplier nodes to surface accounts with comparable concentration risk. Cross-check against the peer-benchmark cohort (47 industrial-equipment manufacturers) to confirm we're seeing a category pattern, not an Acme idiosyncrasy.

Inputs

acme-corp — primary subject
industries.manufacturing — peer cohort filter
bm-122 — benchmark cohort id

[3] code Python · ran 6m ago · 1.84s

# Vector-similarity over supplier-tier-1 to find comparable accounts suppliers = nbrs[nbrs.node_type == "Organization"].node_id.unique() results = p.vector_search( index="acct_embeddings", query="supply chain concentration risk", k=10, filter={"industry": "manufacturing"}, ).to_df()

out DataFrame · 10 rows × 4 cols · 38 ms

rank	account	score	industry
1	acme-corp	0.98	manufacturing
2	globex-industries	0.87	manufacturing
3	initech-holdings	0.81	manufacturing
4	soylent-group	0.76	manufacturing
5	stark-industrial	0.72	manufacturing
6	wayne-mfg	0.68	manufacturing
…	…	…	…

[4] code Python · ran 4m ago · 218ms

# Plot supplier concentration distribution across the peer cohort from egan.benchmarks import cohort bm = cohort("bm-122") fig = bm.distribution(metric="supplier_concentration", highlight="acme-corp") fig.show()

out Figure · matplotlib · 480×220

supplier_concentration · n=47 · highlight: acme-corp

[5] code Python · ran 4m ago · 0.06s

# Hybrid: graph-traverse first, then vector-rank within result set hybrid = p.hybrid_search( graph="MATCH (o:Organization)-[:depends_on]->(s) RETURN s", vector_query="single-source dependency", vector_index="entity_embeddings", rerank_k=20, ).to_df() print(f"Returned {len(hybrid)} hybrid results")

out stdout · 0.06s

✓ Hybrid query OK · graph stage 41ms · vector stage 18ms Returned 20 hybrid results

Add cell

Functions

click to insert

Search functions… ⌘.

Piggiepiggie

cypher(query, **params) ↩ Run Cypher against AGE · returns Result

sql(stmt, *args) ↩ Run raw SQL · returns Result

vector_search(index, query, k=10) ↩ pgvector k-NN over named index

hybrid_search(graph, vector_query) ↩ graph-then-vector or vector-then-graph

connect(env) ↩ PiggieDriver for env (prod, dev, sandbox)

Eganegan

entities.get(id) ↩ Fetch entity by id with properties + edges

entities.find(type, **filters) ↩ Query entities by type and property

taxonomy.industry(name) ↩ Industry hierarchy + sub-industries

compliance.controls(framework) ↩ Controls for framework + mapped activities

benchmarks.cohort(id) ↩ Peer cohort + distribution helpers

accounts.intelligence(id) ↩ Account profile + opportunities + score

Result helpersResult

.to_df() ↩ pandas DataFrame

.to_arrow() ↩ Arrow Table (zero-copy)

.to_dicts() ↩ list[dict]

.graph(expand=True) ↩ Vertex / Edge / Path objects

Data sciencestdlib

pandas, numpy, scipy → Standard scientific Python

scikit-learn, statsmodels → Modeling + statistics

matplotlib, plotly, altair → Visualization (auto-rendered as output)

networkx, graph-tool → Graph algorithms (centrality, communities)

Python 3.13 · running

Memory2.4 GB / 8 GB

Uptime2h 14m

Connectionpiggie · prod

Mobile

Egan

◇ Analyze › Notebook Acme PMP — supply chain

Python 3.13 nb-204 12 cells ran 8m

[1] code 0.42s Run ▶

from piggie import connect p = connect("prod") acme = entities.get("acme-corp") nbrs = p.cypher(...).to_df()

out DataFrame · 247×6

org	rel	node
acme-corp	depends_on	supplier-tier-1
acme-corp	depends_on	logistics-partner
acme-corp	enables	distribution-net
…	…	…

[2] md

Approach
Pull 2-hop neighborhood from Acme, vector-rank suppliers, cross-check against peer cohort to confirm category pattern.

[3] code 1.84s Run ▶

results = p.vector_search( index="acct_embeddings", query="supply chain risk", k=10, ).to_df()

out 10 rows · 38ms

rank	account	score
1	acme-corp	0.98
2	globex-industries	0.87
3	initech-holdings	0.81

Functions · Piggie

cypher(query, **)↩

vector_search(...)↩

hybrid_search(...)↩

Functions · Egan

entities.get(id)↩

taxonomy.industry(name)↩

compliance.controls(...)↩

benchmarks.cohort(id)↩