私のLLM消費の監査：SQLite、OpenCode、Ollama Cloud、そしてなぜ私のプロサブスクリプションが世界市場を破壊するのか

@startuml
skinparam backgroundColor #FEFEFE
skinparam defaultFontSize 12

title 簡略化されたスキーマ — OpenCode SQLite
entity "セッション" as S {
  * id : TEXT
  --
  * model : TEXT (JSON)
  * cost : REAL
  * tokens_input : INTEGER
  * tokens_output : INTEGER
  * tokens_reasoning : INTEGER
  * tokens_cache_read : INTEGER
  * tokens_cache_write : INTEGER
  * time_created : INTEGER
  * project_id : TEXT
}

entity "メッセージ" as M {
  * id : TEXT
  --
  * session_id : TEXT <<FK>>
  * data : TEXT (JSON)
  * time_created : INTEGER
}

entity "プロジェクト" as P {
  * id : TEXT
  --
  * worktree : TEXT
  * name : TEXT
}

S ||--o{ M : "セッションID"
S }o--|| P : "project_id"

note right of S
  La table session agrège
  tokens + coût + modèle.
  C'est notre point d'entrée
  principal.
end note

note right of M
  Les messages individuels.
  Le champ data contient
  le modèle utilisé (format JSON)
  même si la session change
  de modèle en cours de route.
end note

@enduml

{
  "id": "deepseek-v4-pro:cloud",
  "providerID": "ollama-cloud",
  "variant": "default"
}

const { DatabaseSync } = require('node:sqlite');
const db = new DatabaseSync(
  `${process.env.HOME}/.local/share/opencode/opencode.db`,
  { readonly: true }
);

// 1. Statistiques globales
const global = db.prepare(`
  SELECT
    COUNT(*) as sessions,
    SUM(tokens_input) as total_input,
    SUM(tokens_output) as total_output,
    SUM(tokens_reasoning) as total_reasoning,
    SUM(tokens_cache_read) as total_cache_read,
    SUM(cost) as total_cost
  FROM session
`).get();

// 2. Ventilation par modèle et fournisseur
const byModel = db.prepare(`
  SELECT
    json_extract(model, '$.id') as model_id,
    json_extract(model, '$.providerID') as provider,
    COUNT(*) as sessions,
    SUM(tokens_input) as total_input,
    SUM(tokens_output) as total_output,
    SUM(cost) as total_cost
  FROM session
  WHERE model IS NOT NULL
  GROUP BY model_id, provider
  ORDER BY SUM(tokens_input) DESC
`).all();

// 3. Par mois
const monthly = db.prepare(`
  SELECT
    strftime('%Y-%m', datetime(time_created/1000, 'unixepoch')) as month,
    COUNT(*) as sessions,
    SUM(tokens_input) as total_input,
    SUM(tokens_output) as total_output,
    SUM(cost) as total_cost
  FROM session
  WHERE time_created IS NOT NULL
  GROUP BY month
  ORDER BY month
`).all();

// 4. Messages par modèle (détection des changements intra-session)
const msgByModel = db.prepare(`
  SELECT
    json_extract(m.data, '$.model.providerID') as provider,
    json_extract(m.data, '$.model.modelID') as model_id,
    COUNT(*) as messages
  FROM message m
  WHERE json_extract(m.data, '$.model') IS NOT NULL
  GROUP BY provider, model_id
  ORDER BY COUNT(*) DESC
`).all();

console.log({ global, byModel, monthly, msgByModel });
db.close();

@startuml
skinparam backgroundColor #FEFEFE
skinparam defaultFontSize 12

title 月次進化 — トークン入力
rectangle "2026年3月
9セッション
37.7Mトークン" as MARS #FFE082
rectangle "2026年4月
486セッション
13.3億トークン" as AVR #81C784
rectangle "2026年5月
527セッション
12.5億トークン" as MAI #64B5F6

MARS -right-> AVR : ×35
AVR -right-> MAI : stable (période en cours)

note bottom of MAI
  Mai comptabilisé jusqu'au 18.
  Projection mois complet : ~2,1 Md tokens
  si la tendance se maintient.
end note
@enduml

Modèles utilisés dans les messages (ollama-cloud provider) :
├─ deepseek-v4-pro      1 441 msgs  (32%)
├─ qwen3.5:397b         1 217 msgs  (27%)
├─ glm-5.1                712 msgs  (16%)
├─ kimi-k2.6:cloud        634 msgs  (14%)
├─ qwen3-coder:480b       414 msgs   (9%)
├─ gpt-oss:20b             34 msgs  (0,8%)
├─ qwen3-coder-next        27 msgs  (0,6%)
└─ gemma4:31b               7 msgs  (0,2%)