{"id":"knowledge","type":"module","title":"Knowledge (Org Knowledge Graph and Packages)","status":"ga","audience":["consultant","bid-manager","org-admin"],"scope":"both","tier_min":"free","aliases":["8","capability-graph","playbooks"],"summary":"The org-specific context substrate. It turns a pile of PDFs into a structured capability graph with evidence, plus a searchable knowledge corpus and shareable packages the rest of the AI stack reads from.","keywords":["knowledge-graph","capabilities","astaroth","packages","marketplace","cv","pds","mephiston"],"body":{"what_it_is":"Klarum's firm-specific knowledge layer: the system of record for everything a\nfirm *knows and has done* - company profile, project references, project\ndetail sheets (PDS), consultant CVs, past EOIs, collateral - turned from a\npile of PDFs/DOCX into a **structured, queryable capability graph** plus a\n**searchable knowledge corpus.** Three things live here:\n\n1. **Firm Knowledge Graph**: uploaded docs are chunked, embedded, and\n   LLM-tagged with **capabilities** (e.g. `road_safety_audit`,\n   `pavement_management`). Result: a tree\n   `Firm to Category to Capability to evidencing Documents`, each capability\n   carrying a confidence label (`explicit`/`implied`), evidence score,\n   passage counts, and sample quotes.\n2. **Knowledge Items**: curated records (`title` + markdown `body` + a `kind`\n   like standard/whitepaper/ICP/case study), org-scoped and semantically\n   searchable.\n3. **Knowledge Packages**: named, versioned bundles of items, shareable via\n   public link, exportable (ZIP / branded PDF), installable into a workspace\n   or another tenant (\"playbooks\"/marketplace).\n\nIt is the firm's **context substrate**: the matching engine, EOI/document\ngenerator, and workspace agent all draw on it. Without it, those AI features\nrun cold.\n","what_it_helps_users_do":"Make the firm's track record machine-usable (dump 5-10 reference docs to an\nauto-derived capability inventory); power matching with firm-specific signal;\ndraft better EOIs faster (past EOIs/PDS become retrievable, citable context\nwith page provenance); surface deployable personnel (bulk-uploaded CVs to a\nTeam Composition view by role/seniority); audit ingest health (per-document\nstatus, retry/bulk-retry/reingest); package and share institutional\nknowledge; and search the org semantically.\n"},"engines":[],"integrations":[{"id":"astaroth-sidecar","type":"integration","title":"Astaroth capability sidecar","summary":"The graph-construction brain: a doc-level capability compiler plus a per-chunk tagger, both over AWS Bedrock Converse, constrained to the seeded capability vocabulary.","category":"ai-sidecar"}],"capabilities":[{"id":"capability-graph","type":"capability","title":"Firm knowledge graph","summary":"An org to category to capability to evidencing document tree, where every has_capability edge points back at the exact passage, page and confidence that justifies it.","status":"ga","scope":null,"keywords":["graph","edges","evidence","confidence"],"body":{"what_it_is":"The sub-view that is the producer of the capability graph - LlamaParse,\nCohere/bge-m3, pgvector + Mephiston, Astaroth two-stage Bedrock\nextraction, graph edges; status synthesised defensively so the UI never\nshows perpetual spinners.\n","what_it_helps_users_do":null}},{"id":"knowledge-items","type":"capability","title":"Knowledge items","summary":"Curated org-scoped records with a title, markdown body and a kind. A sweep embeds them every 30 minutes, so a new item is keyword-findable at once and vector-findable at the next sweep.","status":"ga","scope":null,"keywords":["curated","semantic-search","items"],"body":{"what_it_is":"Curated records (`title` + markdown `body` + a `kind` like\nstandard/whitepaper/ICP/case study), org-scoped and semantically\nsearchable. The sub-view carries items, org-wide semantic search and the\npackage builder.\n","what_it_helps_users_do":null}},{"id":"knowledge-packages","type":"capability","title":"Knowledge packages","summary":"Named, versioned bundles of items, exportable as ZIP or branded PDF and installable into a workspace or another tenant. Cross-tenant public packages form a marketplace.","status":"ga","scope":null,"keywords":["packages","playbooks","marketplace","install"],"body":{"what_it_is":"Named, versioned bundles of items, shareable via public link, exportable\n(ZIP / branded PDF), installable into a workspace or another tenant\n(\"playbooks\"/marketplace). The build is manual selection.\n","what_it_helps_users_do":null}},{"id":"team-composition","type":"capability","title":"Team composition","summary":"Bulk-uploaded CVs are extracted for job title and seniority and rolled up into a view of who the org can actually deploy.","status":"ga","scope":null,"keywords":["cv","seniority","staffing"],"body":null},{"id":"ingest-health","type":"capability","title":"Ingest health","summary":"Per-document parse and tag status with retry, bulk retry and reingest, and status synthesised defensively so the UI never shows a perpetual spinner.","status":"ga","scope":null,"keywords":["status","retry","observability"],"body":null}]}