mvanhorn OpenRouter — Printing Press CLI
Prerequisites: Install the CLI
This skill drives the openrouter-image-pp-cli binary. You must verify the CLI is installed before invoking any command from this skill. If it is missing, install it first:
- Install via the Printing Press installer. It defaults binaries to
$HOME/.local/binon macOS/Linux and%LOCALAPPDATA%\Programs\PrintingPress\binon Windows:npx -y @mvanhorn/printing-press-library install openrouter-image --cli-only - Verify:
openrouter-image-pp-cli --version - Ensure the reported install directory is on
$PATHfor the agent/runtime that will invoke this skill.
If the npx install fails (no Node, offline, etc.), fall back to a direct Go install (requires Go 1.26.5 or newer). This installs into $GOPATH/bin (default $HOME/go/bin), so add that directory to $PATH instead:
go install github.com/mvanhorn/printing-press-library/library/ai/openrouter-image/cmd/openrouter-image-pp-cli@latest
If --version reports "command not found" after install, the runtime cannot see the binary directory on $PATH. Do not proceed with skill commands until verification succeeds.
OpenRouter's Image API fronts 40+ image models from every major lab. This CLI adds what the API alone lacks: offline model ranking by capability and budget, pre-spend cost estimates, deterministic re-generation from a local history ledger, budget-gated batch runs, and a weekly spend digest. Model selection is always explicit — every generation names its model.
When to Use This CLI
Use this CLI when an AI agent or human needs to generate images on demand through OpenRouter with explicit model selection, wants to pick the cheapest capable provider before spending, run budgeted batches from a CSV, reproduce past generations exactly, or track image-generation spend over time. It is the tool for scheduled image pipelines, prompt iteration across models, and cost-aware image production.
Anti-triggers
Do not use this CLI for:
- Do not use this CLI for chat completions or text generation — use the existing openrouter-pp-cli (cost attribution) or a chat CLI
- Do not use this CLI for video generation — OpenRouter's video API is a different surface
- Do not use this CLI for embedding, reranking, or speech endpoints
- Do not use this CLI as a general HTTP client for arbitrary OpenRouter endpoints
Unique Capabilities
These capabilities aren't available in any other tool for this API.
Local state that compounds
models rank— Rank every image model+provider combo cheapest-first under your capability and budget constraints.Pick the cheapest provider that meets your image constraints without paging through catalog JSON.
openrouter-image-pp-cli models rank --image-to-image --resolution 4K --max-cost 0.10 --limit 5 --jsoncost-estimate— Estimate USD cost of a generation before spending credits, computed offline from synced per-endpoint pricing.Agents can check the price of a planned image before spending credits.
openrouter-image-pp-cli cost-estimate --model openai/gpt-image-1 --resolution 2K --quality high --n 4regenerate— Re-run a past generation with its exact stored parameters (model, seed, resolution, quality, references).Reproduce or tweak a past image without re-typing the full flag set.
openrouter-image-pp-cli regenerate gen-1234567890 --output winner.pngusage digest— Period-over-period spend and volume summary: images generated, USD spent, top models, cost per image vs the prior window.Budget owners get a machine-readable weekly cost report from the local ledger.
openrouter-image-pp-cli usage digest --since 7d --agent
Agent-native plumbing
batch— Run many generations from a CSV with a hard USD budget: estimate first, abort before any spend if over, then execute and log each cost.Cron pipelines can fire a batch with a hard spend cap and get typed exit codes instead of burning the whole balance.
openrouter-image-pp-cli batch --spec batch.csv --budget 2.00 --dry-run
Reachability mitigation
models diff— See newly added, retired, and price-changed image models between syncs so pinned pipelines never break silently.Catch a retired model before the next scheduled batch 404s.
openrouter-image-pp-cli models diff --since 7d --json
Command Reference
activity — Manage activity
openrouter-image-pp-cli activity— Returns user activity data grouped by endpoint for the last 30 (completed) UTC days.
audio — Manage audio
openrouter-image-pp-cli audio create-speech— Synthesizes audio from the input text. Returns a raw audio bytestream in the requested format (e.g. mp3, pcm, wav).openrouter-image-pp-cli audio create-transcriptions— Transcribes audio into text.
benchmarks — Benchmarks endpoints
openrouter-image-pp-cli benchmarks— Unified benchmark endpoint that aggregates scores from multiple benchmark sources (Artificial Analysis, Design Arena
byok — BYOK endpoints
openrouter-image-pp-cli byok create-byokkey— Create a new bring-your-own-key (BYOK) provider credential.openrouter-image-pp-cli byok delete-byokkey— Delete (soft-delete) a bring-your-own-key (BYOK) provider credential by itsid.openrouter-image-pp-cli byok get-byokkey— Get a single bring-your-own-key (BYOK) provider credential by itsid.openrouter-image-pp-cli byok list-byokkeys— List the bring-your-own-key (BYOK) provider credentials for the authenticated entity's default workspace.openrouter-image-pp-cli byok update-byokkey— Update an existing bring-your-own-key (BYOK) provider credential by itsid.
chat — Chat completion endpoints
openrouter-image-pp-cli chat— Sends a request for a model response for the given chat conversation. Supports both streaming and non-streaming modes.
classifications — Task classification market-share endpoints
openrouter-image-pp-cli classifications— Returns the market-share breakdown of OpenRouter traffic by task classification (e.g.
credits — Credit management endpoints
openrouter-image-pp-cli credits create-coinbase-charge— Deprecated.openrouter-image-pp-cli credits get— Get total credits purchased and used for the authenticated user.
datasets — Datasets endpoints
openrouter-image-pp-cli datasets get-app-rankings— Returns the top public apps on OpenRouter ranked by token usage inside the requested date windowopenrouter-image-pp-cli datasets get-rankings-daily— Returns the top 50 public models per day by total token usage on OpenRouter
embeddings — Text embedding endpoints
openrouter-image-pp-cli embeddings create— Submits an embedding request to the embeddings routeropenrouter-image-pp-cli embeddings list-models— Returns a list of all available embeddings models and their properties
endpoints — Endpoint information
openrouter-image-pp-cli endpoints— Preview the impact of ZDR on the available endpoints
files — Files endpoints
openrouter-image-pp-cli files delete— Deletes a file owned by the requesting workspace. Deletion is irreversible.openrouter-image-pp-cli files get-metadata— Retrieves metadata for a single file owned by the requesting workspace.openrouter-image-pp-cli files list— Lists files belonging to the workspace of the authenticating API key.openrouter-image-pp-cli files upload— Uploads a file to be referenced in future API calls.
generation — Generation history endpoints
openrouter-image-pp-cli generation get— Get request & usage metadata for a generationopenrouter-image-pp-cli generation list-content— Get stored prompt and completion content for a generationopenrouter-image-pp-cli generation submit-feedback— Submit structured feedback on a generation the authenticated user made.
guardrails — Guardrails endpoints
openrouter-image-pp-cli guardrails create— Create a new guardrail for the authenticated user.openrouter-image-pp-cli guardrails delete— Delete an existing guardrail. Management key required.openrouter-image-pp-cli guardrails get— Get a single guardrail by ID. Management key required.openrouter-image-pp-cli guardrails list— List all guardrails for the authenticated user.openrouter-image-pp-cli guardrails list-key-assignments— List all API key guardrail assignments for the authenticated user.openrouter-image-pp-cli guardrails list-member-assignments— List all organization member guardrail assignments for the authenticated user.openrouter-image-pp-cli guardrails update— Update an existing guardrail. Collection fields use replace semantics: send the full desired set on every update.
images — Images endpoints
openrouter-image-pp-cli images create— Generates an image from a text prompt via the image generation routeropenrouter-image-pp-cli images list-model-endpoints— Returns the full per-endpoint records for an image model: each endpoint's definitive supported parameters, pricingopenrouter-image-pp-cli images list-models— Lists every image generation model with its top-level supported-parameter superset and a URL to its full per-endpoint
key — Manage key
openrouter-image-pp-cli key— Get information on the API key associated with the current authentication session
keys — Manage keys
openrouter-image-pp-cli keys create— Create a new API key for the authenticated user. The plaintextkeyis returned only in this response.openrouter-image-pp-cli keys delete— Delete an existing API key. Management key required.openrouter-image-pp-cli keys get— Get a single API key by hash. Management key required.openrouter-image-pp-cli keys list— List all API keys for the authenticated user. Management key required.openrouter-image-pp-cli keys update— Update an existing API key. Management key required.
messages — Manage messages
openrouter-image-pp-cli messages— Creates a message using the Anthropic Messages API format. Supports text, images, PDFs, tools, and extended thinking.
model — Model information endpoints
openrouter-image-pp-cli model <author> <slug>— Returns full details for a single model identified by its author and slug (e.g. openai/gpt-4).
models — Model information endpoints
openrouter-image-pp-cli models get— List all models and their propertiesopenrouter-image-pp-cli models list-count— Get total count of available modelsopenrouter-image-pp-cli models list-user— List models filtered by user provider preferences, [privacy settings](https://openrouter.
observability — Observability endpoints
openrouter-image-pp-cli observability create-destination— Create a new observability destination. A maximum of 5 destinations per type is allowed.openrouter-image-pp-cli observability delete-destination— Delete an existing observability destination. This performs a soft delete.openrouter-image-pp-cli observability get-destination— Fetch a single observability destination by its UUID.openrouter-image-pp-cli observability list-destinations— List the observability destinations configured for the authenticated entity's default workspace.openrouter-image-pp-cli observability update-destination— Update an existing observability destination. Only the fields provided in the request body are updated.
openrouter-analytics — Manage openrouter analytics
openrouter-image-pp-cli openrouter-analytics get-meta— Returns the available metrics, dimensions, filter operators, and granularities for the analytics query endpoint.openrouter-image-pp-cli openrouter-analytics query— Execute an analytics query with specified metrics, dimensions, filters, and time range.
openrouter-auth — Manage openrouter auth
openrouter-image-pp-cli openrouter-auth create-keys-code— Create an authorization code for the PKCE flow to generate a user-controlled API keyopenrouter-image-pp-cli openrouter-auth exchange-code-for-apikey— Exchange an authorization code from the PKCE flow for a user-controlled API key
organization — Organization endpoints
openrouter-image-pp-cli organization— List all members of the organization associated with the authenticated management key.
presets — Presets endpoints
openrouter-image-pp-cli presets get— Retrieves a preset by its slug with its currently designated version inline.openrouter-image-pp-cli presets list— Lists all presets for the authenticated user, ordered by most recently updated first.
providers — Provider information endpoints
openrouter-image-pp-cli providers— List all providers
rerank — Rerank endpoints
openrouter-image-pp-cli rerank— Submits a rerank request to the rerank router
responses — OpenAI-compatible Responses API endpoints
openrouter-image-pp-cli responses create— Creates a streaming or non-streaming response using OpenResponses API formatopenrouter-image-pp-cli responses create-compact— Rewrites a conversation into a smaller context window, returning the canonical next context window
scim — SCIM endpoints
openrouter-image-pp-cli scim create-group-mapping— Create a SCIM group-to-workspace role mapping.openrouter-image-pp-cli scim delete-group-mapping— Delete a SCIM group-to-workspace mapping. Management key required.openrouter-image-pp-cli scim get-group-mapping— Get a SCIM group-to-workspace mapping. Management key required.openrouter-image-pp-cli scim list-group-mappings— List SCIM group-to-workspace mappings for the organization.openrouter-image-pp-cli scim list-groups— List SCIM groups for the organization. Management key required.openrouter-image-pp-cli scim update-group-mapping— Update a SCIM group mapping role. Management key required.
videos — Manage videos
openrouter-image-pp-cli videos create— Submits a video generation request and returns a polling URL to check statusopenrouter-image-pp-cli videos get— Returns job status and content URLs when completedopenrouter-image-pp-cli videos list-models— Returns a list of all available video generation models and their properties
workspaces — Workspaces endpoints
openrouter-image-pp-cli workspaces create— Create a new workspace for the authenticated user.openrouter-image-pp-cli workspaces delete— Delete an existing workspace. The default workspace cannot be deleted.openrouter-image-pp-cli workspaces get— Get a single workspace by ID or slug. Management key required.openrouter-image-pp-cli workspaces list— List all workspaces for the authenticated user.openrouter-image-pp-cli workspaces update— Update an existing workspace by ID or slug. Management key required.
Freshness Contract
This printed CLI owns bounded freshness only for registered store-backed read command paths. In --data-source auto mode, those paths check sync_state and may run a bounded refresh before reading local data. --data-source local never refreshes. --data-source live reads the API and does not mutate the local store. Set OPENROUTER_IMAGE_NO_AUTO_REFRESH=1 to skip the freshness hook without changing source selection.
Covered paths:
openrouter-image-pp-cli activityopenrouter-image-pp-cli activity getopenrouter-image-pp-cli activity listopenrouter-image-pp-cli activity searchopenrouter-image-pp-cli benchmarksopenrouter-image-pp-cli benchmarks getopenrouter-image-pp-cli benchmarks listopenrouter-image-pp-cli benchmarks searchopenrouter-image-pp-cli byokopenrouter-image-pp-cli byok getopenrouter-image-pp-cli byok listopenrouter-image-pp-cli byok searchopenrouter-image-pp-cli datasetsopenrouter-image-pp-cli datasets getopenrouter-image-pp-cli datasets listopenrouter-image-pp-cli datasets searchopenrouter-image-pp-cli datasets-rankings-dailyopenrouter-image-pp-cli datasets-rankings-daily getopenrouter-image-pp-cli datasets-rankings-daily listopenrouter-image-pp-cli datasets-rankings-daily searchopenrouter-image-pp-cli embeddingsopenrouter-image-pp-cli embeddings getopenrouter-image-pp-cli embeddings listopenrouter-image-pp-cli embeddings searchopenrouter-image-pp-cli endpointsopenrouter-image-pp-cli endpoints getopenrouter-image-pp-cli endpoints listopenrouter-image-pp-cli endpoints searchopenrouter-image-pp-cli filesopenrouter-image-pp-cli files getopenrouter-image-pp-cli files listopenrouter-image-pp-cli files searchopenrouter-image-pp-cli generationopenrouter-image-pp-cli generation getopenrouter-image-pp-cli generation listopenrouter-image-pp-cli generation searchopenrouter-image-pp-cli guardrailsopenrouter-image-pp-cli guardrails getopenrouter-image-pp-cli guardrails listopenrouter-image-pp-cli guardrails searchopenrouter-image-pp-cli guardrails-assignments-keysopenrouter-image-pp-cli guardrails-assignments-keys getopenrouter-image-pp-cli guardrails-assignments-keys listopenrouter-image-pp-cli guardrails-assignments-keys searchopenrouter-image-pp-cli guardrails-assignments-membersopenrouter-image-pp-cli guardrails-assignments-members getopenrouter-image-pp-cli guardrails-assignments-members listopenrouter-image-pp-cli guardrails-assignments-members searchopenrouter-image-pp-cli imagesopenrouter-image-pp-cli images getopenrouter-image-pp-cli images listopenrouter-image-pp-cli images searchopenrouter-image-pp-cli keysopenrouter-image-pp-cli keys getopenrouter-image-pp-cli keys listopenrouter-image-pp-cli keys searchopenrouter-image-pp-cli modelsopenrouter-image-pp-cli models getopenrouter-image-pp-cli models listopenrouter-image-pp-cli models searchopenrouter-image-pp-cli models-countopenrouter-image-pp-cli models-count getopenrouter-image-pp-cli models-count listopenrouter-image-pp-cli models-count searchopenrouter-image-pp-cli models-useropenrouter-image-pp-cli models-user getopenrouter-image-pp-cli models-user listopenrouter-image-pp-cli models-user searchopenrouter-image-pp-cli observabilityopenrouter-image-pp-cli observability getopenrouter-image-pp-cli observability listopenrouter-image-pp-cli observability searchopenrouter-image-pp-cli organizationopenrouter-image-pp-cli organization getopenrouter-image-pp-cli organization listopenrouter-image-pp-cli organization searchopenrouter-image-pp-cli presetsopenrouter-image-pp-cli presets getopenrouter-image-pp-cli presets listopenrouter-image-pp-cli presets searchopenrouter-image-pp-cli providersopenrouter-image-pp-cli providers getopenrouter-image-pp-cli providers listopenrouter-image-pp-cli providers searchopenrouter-image-pp-cli scimopenrouter-image-pp-cli scim getopenrouter-image-pp-cli scim listopenrouter-image-pp-cli scim searchopenrouter-image-pp-cli scim-group-mappingsopenrouter-image-pp-cli scim-group-mappings getopenrouter-image-pp-cli scim-group-mappings listopenrouter-image-pp-cli scim-group-mappings searchopenrouter-image-pp-cli videosopenrouter-image-pp-cli videos getopenrouter-image-pp-cli videos listopenrouter-image-pp-cli videos searchopenrouter-image-pp-cli workspacesopenrouter-image-pp-cli workspaces getopenrouter-image-pp-cli workspaces listopenrouter-image-pp-cli workspaces search
When JSON output uses the generated provenance envelope, freshness metadata appears at meta.freshness. Treat it as current-cache freshness for the covered command path, not a guarantee of complete historical backfill or API-specific enrichment.
Finding the right command
When you know what you want to do but not which command does it, ask the CLI directly:
openrouter-image-pp-cli which "<capability in your own words>"
which resolves a natural-language capability query to the best matching command from this CLI's curated feature index. Exit code 0 means at least one match; exit code 2 means no confident match — fall back to --help or use a narrower query.
Recipes
Cheapest image-to-image model under budget
openrouter-image-pp-cli models rank --image-to-image --max-cost 0.10 --limit 3 --json
Finds capable providers under a per-image budget, cheapest first
Budget-gated batch from CSV
openrouter-image-pp-cli batch --spec batch.csv --budget 5.00 --dry-run
Dry-run estimates every row and aborts if the total exceeds the budget before any spend
Reproduce last week's winner
openrouter-image-pp-cli regenerate gen-1234567890 --output winner-v2.png
Replays the exact stored model, seed, resolution, and quality of a past generation
Pre-flight cost check for an agent
openrouter-image-pp-cli cost-estimate --model bytedance-seed/seedream-4.5 --resolution 2K --n 4 --json
Agents can gate generation on the quoted price before spending credits
Narrow generation output for agents
openrouter-image-pp-cli generate --model google/gemini-2.5-flash-image --prompt 'a red panda astronaut floating in space, studio lighting' --json --agent --select data.0.media_type,usage.cost
Deeply nested generation responses collapse to the fields an agent needs
Spot a retiring model before cron breaks
openrouter-image-pp-cli models diff --since 7d --json
Surfaces retired and price-changed models between syncs so pinned pipelines fail loudly, not silently
Auth Setup
Set OPENROUTER_API_KEY in your environment. The key is read per command; nothing is persisted to disk. Run openrouter-image-pp-cli doctor to verify setup.
Run openrouter-image-pp-cli doctor to verify setup.
Agent Mode
Add --agent to any command. Expands to: --json --compact --no-input --no-color --yes.
Pipeable — JSON on stdout, errors on stderr
Filterable —
--selectkeeps a subset of fields. Dotted paths descend into nested structures; arrays traverse element-wise. Critical for keeping context small on verbose APIs:openrouter-image-pp-cli benchmarks --agent --select id,name,statusPreviewable —
--dry-runshows the request without sendingOffline-friendly — sync/search commands can use the local SQLite store when available
Non-interactive — never prompts, every input is a flag
Explicit retries — use
--idempotentonly when an already-existing create should count as success, and use--ignore-missingonly when a missing delete target should count as success
Response envelope
Commands that read from the local store or the API wrap output in a provenance envelope:
{
"meta": {"source": "live" | "local", "synced_at": "...", "reason": "..."},
"results": <data>
}
Parse .results for data and .meta.source to know whether it's live or local. A human-readable N results (live) summary is printed to stderr only when stdout is a terminal AND no machine-format flag (--json, --csv, --compact, --quiet, --plain, --select) is set — piped/agent consumers and explicit-format runs get pure JSON on stdout.
Paths and state
Agents should treat the CLI's path resolver as part of the runtime contract:
Use
--home <dir>for one invocation, or setOPENROUTER_IMAGE_HOME=<dir>to relocate all four path kinds under one root.Use per-kind env vars only when a specific kind must diverge:
OPENROUTER_IMAGE_CONFIG_DIR,OPENROUTER_IMAGE_DATA_DIR,OPENROUTER_IMAGE_STATE_DIR,OPENROUTER_IMAGE_CACHE_DIR.Resolution order is per-kind env var,
--home,OPENROUTER_IMAGE_HOME, XDG (XDG_CONFIG_HOME,XDG_DATA_HOME,XDG_STATE_HOME,XDG_CACHE_HOME), then platform defaults.configcontains settings likeconfig.tomland profiles.datacontainscredentials.toml,data.db, cookies, and auth sidecars.statecontains persisted queries, jobs, andteach.log.cachecontains regenerable HTTP/cache files.Stored secrets live in
credentials.tomlunder the data dir. Existing legacyconfig.tomlsecrets are read for compatibility and leaveconfig.tomlon the first auth write.Run
openrouter-image-pp-cli doctor --fail-on warnto surface path and credential-location warnings.agent-contextexposes a schema v4pathsblock for agents that need the resolved dirs.For MCP, pass relocation through the MCP host config. The MCP binary does not inherit CLI flags:
{ "mcpServers": { "openrouter-image": { "command": "openrouter-image-pp-mcp", "env": { "OPENROUTER_IMAGE_HOME": "/srv/openrouter-image" } } } }
Fleet precedence: an inherited per-kind env var overrides an explicit --home for that kind. Use OPENROUTER_IMAGE_HOME or per-kind vars as durable fleet levers, and use --home only for a single invocation. Relocation is not reversible by unsetting env vars; move files manually before clearing OPENROUTER_IMAGE_HOME, or doctor will not find credentials left under the former root.
Automatic learning
This CLI ships a self-capturing learning loop. The CLI does its own bookkeeping: every invocation is journaled locally, a failed flag followed by a corrected retry auto-derives a flag_alias candidate, and a teach on a query family without a playbook auto-synthesizes a playbook_candidate from the session's journal. Your job is judgment only: recall first, act on surfaced candidates, teach the final answer, playbook amend when you observe a correction. You never record failures by hand.
Step 1: recall before any discovery
Before list/search/drill commands on a new user question, run:
openrouter-image-pp-cli recall "<user's question>" --agent
The response envelope:
{
"query": "...",
"normalized": "<normalized form>",
"query_entities": ["..."],
"found": true | false,
"match_score": 0.0,
"results": [
{ "resource_id": "...", "resource_type": "...", "venue": "...",
"confidence": 2, "entity_match": "exact|partial|unknown",
"source": "taught|preseed|pattern", "warnings": ["..."] }
],
"mismatches": [ /* only when --debug-mismatches */ ],
"warnings": [ /* top-level */ ],
"candidates": [
{ "id": 12, "class": "flag_alias | playbook_candidate",
"summary": "...", "sightings": 3, "last_seen": "...",
"rationale": "...",
"next_action": ["<trial command>", "openrouter-image-pp-cli learnings confirm 12"] }
],
"playbook": {
"query_family": "...",
"playbook": {
"steps": [ { "cmd": "<command with {slot} substitution>", "purpose": "..." } ],
"entity_slots": ["$ENTITY"],
"expected_tool_calls": 3
},
"slots_resolved": { "$ENTITY": { "token": "<live token>", "canonical": "<canonical>" } },
"notes": "<workarounds + gotchas for this query family>"
},
"notes": "<duplicate surface for non-playbook callers>"
}
Empty-store short-circuit: if the store has no learnings, playbooks, or candidates yet (recall finds nothing and learnings list and learnings candidates are both empty), skip recall for the rest of this session instead of taxing every query; resume recall-first once something has been taught.
Step 2: decision tree
Read candidates, playbook, notes, results[0], and warnings in that order:
if Candidates present (warnings include "candidates_present"):
-> candidates are try-then-confirm, never facts. Follow each candidate's
two-step next_action verbatim: run the trial command first, then run
`learnings confirm <id>` only after the trial verified the behavior.
Reject a wrong candidate with `learnings reject <id>`.
-> NEVER re-teach something recall surfaced as a candidate; confirm or
reject that candidate instead of teaching a duplicate.
-> candidates ride alongside playbooks and resource hits, not instead of
them; continue with the branches below after acting on them.
if Playbook present:
-> READ Playbook.notes verbatim FIRST (workarounds + gotchas the CLI surface doesn't expose)
-> replay Playbook.steps in order, substituting Playbook.slots_resolved entries
for the entity slot tokens. If a step's slot is unresolved, fall back to
discovery for that step only.
-> the Playbook's expected_tool_calls is a budget; if you find yourself running
materially more, record the divergence via `openrouter-image-pp-cli playbook amend`
at end-of-session.
elif Notes present (no Playbook):
-> read Notes verbatim before any discovery step; they carry known gotchas
for this query family even when no structured choreography exists yet.
elif Found AND Results[0].EntityMatch == "exact" AND Results[0].Confidence >= 2:
-> skip discovery; fetch live data for Results[*].ResourceID in parallel
elif Found AND Results[0].EntityMatch == "partial":
-> candidate hint, NOT a hit; read the resource title to validate before trusting
elif (any row in Mismatches[] when --debug-mismatches was passed):
-> treat as cold start; the stored learning is for a different entity
(different canonical resolved from query_entities)
else: // Found == false, no playbook, no notes
-> cold start; run discovery normally; teach the answer afterward (Step 4).
If the family has no playbook yet, that teach auto-synthesizes a
playbook candidate from this session's journal - you do not need to
record one by hand.
Playbook and Notes are orthogonal to the per-resource path. A recall response can carry both a Playbook AND a Results[] hit - use both: the Playbook tells you which choreography to run; the resource hits short-circuit specific steps. Default to skipping mismatches; pass --debug-mismatches only when investigating cold-start surprises.
Candidate judgment details: learnings confirm <id> prints the candidate's full payload before materializing it - check that the printed payload matches the behavior you verified. learnings reject <id> tombstones the derivation signature so the same candidate does not resurface. The envelope carries only the few candidates worth acting on now; openrouter-image-pp-cli learnings candidates lists the full open set.
Graceful degradation: if learnings confirm is an unknown command, you are driving an older binary - ignore the candidates guidance and follow the rest of the protocol.
Step 3: always read warnings
low_confidence: row exists atconfidence<2. Treat as a hint, not a skip-discovery hit.resource_not_in_store: the local store doesn't have the resource the learning points at. The match validator couldn't classify entities — direct-fetch and re-evaluate.cross_alias_match(per-result): the row was taught under a different alias and matched the live query's canonical viaentity_lookups(e.g., a "USA" teach satisfying a "United States" recall). Trust the resource_id.similar_shape_different_entity:<canonical>(top-level): a structurally matching row exists but its canonical entity differs from the live query's. Treated as cold start; the warning carries the conflicting canonical as a hint, but the row is NOT promoted into Results.ambiguous_alias(top-level): a single query entity resolved to multiple canonicals (e.g., "Cards" → Arizona Cardinals + St. Louis Cardinals). Surface the ambiguity from context before committing to a resource.candidates_present(top-level): the envelope carries acandidatessection. Handle it via the candidates branch in Step 2 before anything else.lookup_refresh_available(top-level): an entity in the query has no lookup row yet, but synced data could provide one. Runopenrouter-image-pp-cli syncto refresh entity lookups.- Top-level
no_learnings_for_query_family: the table had no rows above the Jaccard floor. Pure cold start.
Step 4: teach & after finalizing your response - always
Teaching is unconditional. After resolving a query the store could not answer, background-teach the final resource mapping - no call-count threshold, no judging whether it was "worth" learning. The teach is the anchor of the loop: it triggers playbook synthesis for a family without a playbook, and same-referent phrasings fold into one family so near-duplicate teaches do not fragment the store. Fire it after assembling your user-facing response but BEFORE emitting it, with a shell & so the call returns immediately:
openrouter-image-pp-cli teach --query "<user's question>" --resource-type <type> --resource <id1> --resource <id2>
# (append shell `&` to background it)
Silent on success. Errors only land in teach.log under the resolved state dir. Teach the most specific resource - if the user asked a broad question and you walked through parent records to find the specific answer, teach the leaf id, not the parent. The CLI uses seeded entity_lookups for cross-alias resolution at recall time, so a teach under one alias (e.g., "Niners") satisfies future queries under another alias (e.g., "49ers", "San Francisco") automatically.
PII rule: teach the structural question with identifiers stripped - never include names, emails, phone numbers, account ids, or other personal identifiers in taught queries or notes. The CLI scans teach queries for obvious email/phone shapes and warns, but does not block; strip before teaching rather than relying on the warning.
Step 5: playbooks - optional flags, automatic synthesis
You do not need to decide whether a session "deserves" a playbook: a teach on a family without one auto-synthesizes a playbook_candidate from the session's journal, and the next session judges it via confirm/reject. Attach explicit playbook flags only when you already hold choreography worth recording verbatim - workarounds the CLI didn't surface (silently-dropped flags, undocumented params, pagination tricks, payload gotchas). Prefer the integrated one-call form - record the resource learning and the playbook in the same teach invocation:
# Common case: record both the resource learning AND the playbook in one call.
openrouter-image-pp-cli teach \
--query "<user's question>" \
--resource <id> \
--playbook-file ~/playbooks/<shape>.json \
--playbook-notes-file ~/playbooks/<shape>-notes.md
# (append shell `&` to background it)
# Alternate: playbook-only (no resource to record alongside).
openrouter-image-pp-cli teach-playbook \
--query "<user's question>" \
--playbook-file ~/playbooks/<shape>.json \
--notes-file ~/playbooks/<shape>-notes.md
Playbook files are JSON with steps, entity_slots, expected_tool_calls. Notes files are markdown carrying the gotchas verbatim. File-free callers (MCP-only agents) pass the same content inline: --playbook-json and --playbook-notes on the integrated teach form, --playbook-json and --notes on teach-playbook. On the integrated teach form, the playbook flags are optional - omit them entirely for a resource-only teach. On the standalone teach-playbook form, at least one of the playbook and notes flags must be set; both empty is rejected. Playbooks are keyed on the structural query family (entities stripped) so a recipe taught from one entity-shaped query applies to every other query of the same shape, with slots_resolved binding the live query's canonical at recall time.
When you DO find a playbook on a future recall, treat it as ground truth: replay the steps with slots_resolved substitutions, skip the discovery that the choreography already documents, and read notes before any step.
Step 6: playbook amend & when your debug response identifies a correction
If your debug-protocol response identifies a concrete correction the notes or playbook should know — a workaround, an undocumented endpoint shape, a stale field name, observed schema drift, an empty-payload fallback — fire playbook amend BEFORE emitting your user-facing response. Same fire-and-forget posture as teach.
openrouter-image-pp-cli playbook amend \
--query "<exact recall query string>" \
--add-note "<your concrete correction>"
# (append shell `&` to background it)
What counts as worth amending: a behavior you OBSERVED this session that future-you would benefit from knowing. Examples worth amending:
- A workaround for a CLI surface that silently drops or misorders a flag.
- An undocumented endpoint shape (response wrapped in
{meta, results}, payload nested two levels deeper than the docs claim). - Observed schema drift (a field renamed, an index that shifted between seasons, a category label that the API now returns lower-cased).
What does NOT belong in notes:
- The year-specific or entity-specific answer to the user's question. That's the response, not a learning.
- Per-team / per-athlete / per-row data the playbook already retrieves at runtime.
- Statements that paraphrase what the existing notes already say.
The amend command appends to the family's existing notes with a timestamped marker ([amend YYYY-MM-DDTHH:MMZ]: <text>). Multiple amends accumulate; the audit trail is visible. If no playbook exists yet for the family, amend creates a notes-only one (so cold-start corrections still land).
PII discipline for amend notes
playbook amend notes are designed to potentially flow upstream as shared knowledge in future versions of the Printing Press. Keep them clean of user-identifying content so the upstream-contribution path stays open without retroactive scrubbing:
- Do NOT embed paths to user filesystems, personal API keys or tokens, user email addresses, user GitHub handles, or specific query histories tied to a single user.
- Acceptable: endpoint shapes, undocumented field names, API gotchas, observed schema drift, workarounds for CLI surfaces, generalizable pagination or retry tactics.
If a correction is only meaningful with user-specific context, it belongs in a personal note, not in the playbook amend.
Measuring the loop
openrouter-image-pp-cli learnings stats reports recall hit rate, teach-to-reuse, playbook resolution rate, and candidate confirm/reject counts from the local learn_events table. Rates are null until they have a denominator; everything stays on this machine. Use it to check whether the loop is earning its keep for this CLI.
Disabling learning
--no-learnon a single command short-circuits bothrecalland theteachwrite path. Use for deterministic agent flows or tests that must not be affected by accumulated learnings.OPENROUTER_IMAGE_NO_LEARN=truein the environment globally disables the pipeline.
Agent Feedback
When you (or the agent) notice something off about this CLI, record it:
openrouter-image-pp-cli feedback "the --since flag is inclusive but docs say exclusive"
openrouter-image-pp-cli feedback --stdin < notes.txt
openrouter-image-pp-cli feedback list --json --limit 10
En
…(truncated)