Trustbase rCB Data Schemav1.9 · 2026-09-20

The chemical recycling (pyrolysis) industry has no public, structured, machine-readable data source that an AI agent can rely on. Trustbase Lab fills that gap — and further anchors every rCB record against a native carbon black benchmark layer so that substitution economics and technical feasibility can be compared like-for-like.

• Patch (1.0 → 1.1): additive optional fields, documentation, no breaking changes.
• Minor (1.x → 1.y): additive modules, new ID prefixes, content growth — no breaking change to existing fields.
• Major (1.x → 2.0): new required fields, removed deprecated fields, breaking change. Announced 30 days before promotion.
• The current version is v1.9 (expanded equipment vendors to the pyrolysis family — multi-feedstock tyre / plastic / sludge / biomass). Schema version and MCP tool count move independently — the "27 → 30" figure here and in the version history below is the change introduced by v1.9 at the time, not the live tool count (live count: see the MCP integration section).
• Data files MUST include meta.schema_version. Consumers MUST pin to a major version.

{
  "meta":                  { /* dataset-level metadata, schema_version, counts, verified_scheme */ },
  "domestic":              [ /* 37 Chinese companies / projects           (v1.0) */ ],
  "international":         [ /* 35 international companies                (v1.0) */ ],
  "vendors":               [ /* 58 equipment vendors                       (v1.3) */ ],
  "policies":              [ /* 49 policy & regulation records             (v1.8) */ ],
  "techs":                 [ /* 44 technical standards                     (v1.8) */ ],
  "articles":              [ /* 64 articles (first-hand research from the "再生炭黑产业" WeChat account)  (v1.4) */ ],
  "grades":                [ /* 37 native / conductive / rCB grades        (v1.5) */ ],
  "knowledge_cards":       [ /* 61 public knowledge cards                  (v1.7) */ ],
  "price_trends":          { /* trend direction + official anchors only, no prices  (v1.7) */ },
  "substitution_boundaries":{/* 4-dim substitution boundaries               (v1.7) */ },
  "rcb_index":             { /* rCB price index (anchor factor × value factor, no historical fill)  (v1.7) */ },
  "trade_flows":           { /* 1751 UN Comtrade HS 280300 records          (v1.8) */ }
}

Total current count: 2,348 records as of data_version: 0.12. The dataset is a mix of list-of-record tables (companies / policies / techs / articles / grades / knowledge_cards / trade_flows) and dict modules (price_trends / substitution_boundaries / rcb_index) — see §4.4 for the field contract per asset type.

{
  "id":          "rcb-cn-008",
  "public":      { /* free-to-display fields */ },
  "gated":       { /* detailed facts; phones desensitised */ },
  "description": { "zh": "...", "en": "..." },
  "as_of":       "2026-06",
  "type":        "native",   // v1.5 only — grades module: native / rcb / conductive
  "key":         "N550"      // v1.5 only — semantic key for grades
}

The most important fields for AI agents. Pre-shaped, structured summaries designed to be fed directly into an LLM prompt without further parsing. Both fields are required (P0-6).

{
    "id":          "cb-trade-chn-764-2024-x",
    "reporter":     {"iso3": "CHN", "code": 156, "name": "China"},
    "partner":      {"iso3": "THA", "code": 764, "name": "泰国"},
    "hs_code":      "280300",
    "period":       "2024",
    "flow":         "X",
    "quantity_t":   270184.47,
    "value_usd":    337500001.0,
    "unit_price_usd_t": 1249.42,
    "share_pct":    28.07,
    "source":       {"org": "UN Comtrade", "dataset": "API v1 / C/A/HS", "as_of": "2024"},
    "confidence":   "A"
  }

Stable IDs (rcb-cn/gl/eq/pol/tech/art-NNN) are persisted in id_map.json — adding a new entity never changes an existing ID. Semantic IDs (rcb-grade-*, cb-trade-*, CB-…) are derived from authoritative keys and locked.

AI trust infrastructure. Agents auto-deciding procurement partners should filter on verified=true or call the MCP tool get_verified_suppliers.

{
  "public": { "verified": true },
  "gated":  { "confidence": "A" }
}

Rule: verified=true ⇔ confidence=A. Enforced by data_steward (P0-10).

Breakdown (company database): 21 Chinese + 16 international + 22 equipment vendors = 59 companies verified. Cross-corpus A-grade count (companies + vendors + policies + techs + articles + grades) = 223 / 324 = 68.8%.

• Every record carries native (name, location, status) AND English (name_en) content.
• Two parallel description blocks are provided.
• Default output language is zh-CN (Chinese). English via language="en" on MCP tools.
• Build engine: en_lexicon.py (dictionary-greedy longest match); zero pinyin fallback.

MCP server at https://trustbaselab.com/mcp exposes 57 tools. All tools support response_format='json'/'markdown' and accept language='zh'/'en'.

curl -X POST https://trustbaselab.com/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"get_verified_suppliers",
                 "arguments":{"language":"en","response_format":"json"}}}'

curl -X POST https://trustbaselab.com/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call",
       "params":{"name":"query_trade_flows",
                 "arguments":{"reporter":"CHN","period":"2024","flow":"X",
                              "response_format":"json","limit":50}}}'

Returns 119 CHN 2024 export records; quantity_t already converted from kg (P0-15 guard).

curl -X POST https://trustbaselab.com/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call",
       "params":{"name":"get_trade_partner_ranking",
                 "arguments":{"reporter":"CHN","period":"2024","flow":"X",
                              "top_n":5,"response_format":"json"}}}'

Returns Top-5 with quantity_t, value_usd, unit_price_usd_t, share_pct. Sample first record for CHN 2024 X: Thailand ≈ 270,184 t (28.07% of total exports).

• Source-of-truth: a maintained master workbook plus _src/*.json module files.
• Pipeline (v1.7+): _src/ → export_json.py (merge + assign stable IDs + SCRUB_RULES) → data_steward.py (14 P0 checks: P0-1..10 base, P0-11 negative list, P0-12 red-line wording, P0-13 confidentiality level, P0-14 knowledge-base gate, P0-15 trade-table specific) → build_ai_layer.py → upload.
• Trade flows (v1.8): _tools/fetch_comtrade.py → _src/trade_flows.json (merged from CHN 157 + GLOBAL 1449 + World 145 → 1751 records) → export_json.py → P0-15. Re-fetchable on demand with the user's UN Comtrade subscription key.
• Every record is assigned a stable entity ID via id_map.json — never reused or renumbered.
• Each export masks all 11-digit Chinese mobile numbers before publication.
• SCRUB_RULES (in export_json.py) sweeps internal names and operator-entity mentions out before write — pipeline-level and non-bypassable.
• Re-export cadence: weekly inspection; urgent refresh on industry events.

When citing this schema in academic or commercial work: