Two opt-in envs preparing the openclaw supplier on inference-proxy:
OPENAI_SESSION_PASSTHROUGH=1 — chat-session turns speak the stateful-
upstream contract: the escrow ref is sent as the OpenAI `user` field
(OpenClaw keys one persistent agent session per `user`, so browser/
tool state survives across turns) and only the current turn's delta
goes upstream — the upstream session carries history, and re-sending
the full transcript would duplicate it into the upstream context every
turn (OpenClaw renders ALL request messages into the new agent turn).
The local transcript still accumulates fully for the receipt. Without
`user`, OpenClaw creates an ephemeral session per request (uuid key),
so default-off stays safe for stateless providers.
OPENAI_MODEL_OVERRIDE — fixed upstream model id sent instead of the
advert's buyer-facing model. OpenClaw's chat-completions endpoint 400s
on anything but "openclaw" / "openclaw/<agentId>", while the advert
must carry a unique fleet-routable id (openclaw-web-agent).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Eight new mainnet suppliers backed by Hetzner's OpenAI-compatible
inference API (https://inference.hetzner.com/api). Pure configuration —
the LLM_BACKEND=openai path is reused unchanged, following the
HuggingFace-router precedent:
kimi-code[-chat] Kimi-K2.7-Code (ctx 262144)
glm[-chat] GLM-5.2-NVFP4 (ctx 512000)
ds-flash-htz[-chat] DeepSeek-V4-Flash-0731 (ctx 512000)
qwen35b[-chat] Qwen/Qwen3.6-35B-A3B-FP8 (ctx 262144)
Uniform 0.2 AP3X per job, default 1 AP3X bonds, no upstream max_tokens
(advert cap = model context length), 300s SLA (1800s for chat sessions).
All four models smoke-tested for non-empty content and tools round-trip.
Runbook: docs/HETZNER_SETUP.md. Adverts, wallets, and host env files are
operator steps after merge; first `up -d` is manual (CD skips projects
with no running containers).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fixes 2 critical, 2 important, and 5 minor findings from review of 069edcb:
- C1: add a happy-path test that reaches postOcrEscrow.ts's escrowLockFloor
call and asserts the locked amount; verified it fails against a bare
economicTotal and passes against the real code.
- C2: gate /v1/ocr/extract (auth + rate limits) before the 16mb body parser
so an unauthenticated caller can no longer force a full buffer/JSON.parse.
- I2: hoist result/used/escrowRef in the OCR one-shot handler so the failure
path records escrow_ref/supplier/cost when an escrow was actually posted,
and emit an ocr_failed structured log line mirroring ocr_settled.
- I3: export venueAttribution and cover its sanitization directly; add a
stubbed happy-path route test asserting the first-party-brokered
disclosure label reaches the response.
- Minor: drop unused hfWeightsSource end to end, rename a misleading test,
make the OCR capability template's pinning language conditional on the
self-hosted upstream, tighten the /v1/ocr path-prefix guard, and make
preflight's required_lovelace use the same escrowLockFloor the builder
locks (exported from the shared tx barrel for gateway use).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Adds the ocr.page.extract.chandra-ocr-2.v1 capability end to end: the
postOcrEscrow builder (locking the min-ada floor for the Submitted-state
datum, required since the min-ada fix), the gateway OCR route that selects
an OCR supplier by capability, posts a per-page escrow, calls the supplier,
verifies and settles, and returns content plus the signed receipt labelled
first-party-brokered. Supplier OCR handler with a hosted document-API
upstream and buyer-SDK OCR support included.
Squashed from the 2026-08-06 working subset; the deploy pack is
intentionally excluded (it lives in a private operator repo).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
OpenAI-compatible compute-backend preset (config only, reuses LLM_BACKEND=openai) for pointing a supplier node at the HuggingFace Inference Providers router. Adds a boot guard that rejects the OpenRouter-only reasoning param against the HF router, unit tests, and an operator runbook.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The lockfile gained patchedDependencies (libsodium-wrappers-sumo) in
2aa1688, and the package manager hashes the patch file during install —
every image build since then fails with ENOENT on /repo/patches. No
image had been rebuilt since, so it surfaced only on the next deploy.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
planConsolidate returned already-healthy only for wallets with EXACTLY
2 UTxOs, but buildConsolidateWalletTx lands 3 (collateral + working +
the fee-reserve remainder as change). The shape consolidation produces
can never satisfy its own health check, so every wallet-health tick
re-consolidates and pays the ~0.5 AP3X min fee again. Measured in a
production deployment: a 45s ticker submitted 232 consolidations and
burned ~116 AP3X in about 2.5 idle hours; at the default 600s interval
the same loop leaks ~3 AP3X per hour whenever the wallet is not exactly
2 UTxOs.
Healthy is now structural: a pure-lovelace UTxO >= collateralLovelace
exists, at least 1 AP3X sits outside it, and fragmentation is bounded
(<= 6 UTxOs, new optional maxUtxos param). The health check also runs
BEFORE the low-balance guard, so a wallet that can already run script
transactions is never reshaped or refused. walletHealth gains two
independent circuit breakers - halt after 3 consecutive consolidations
that never reach healthy, halt past 6 consolidations per 24h - so any
future non-convergence costs at most ~1.5 AP3X instead of the whole
wallet, plus a docstring that matches reality.
Tests: the exact post-consolidation-shape regression, one-step
idempotence against the builder output model, candidate boundaries
(native assets, 1,628-lovelace-short), healthy-below-reshape-threshold,
and breaker behavior including error and reset paths.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two structural fixes for demo concurrency (3 single-slot suppliers thrashed
under Open WebUI's ~3 conversations per message):
Multi-slot suppliers (MAX_CHAT_SESSIONS, default 1):
- SupplierState is now capacity-aware: a Set of active escrow refs with
free-until-full status semantics, ref-scoped release(ref) (also fixes the
latent bug where handler catch-alls released another request's slot), and
a walletMutex serializing every wallet-spending chain op — Claim, Submit,
and wallet-health consolidation queue instead of racing the single
working/collateral UTxO. /status gains active_sessions/max_sessions.
Entry-ticket settle mode (CHAT_SETTLE_MODE=ticket, default full, both sides):
- The buyer's escrow post at session open is the ONLY chain op. The supplier
verifies the Open escrow as an entry ticket but never Claims/Submits
(/v1/chat/start → {status:"ticket"}; /v1/chat/end → receipt-less
{status:"ended"}); no hard-cap timer, so sessions may outlive deliver_by.
Used tickets can't reopen sessions (records retained; 409 session_exists).
- Buyer SDK: startChat/endChat are settle-mode aware, driven by the
supplier's response — mixed-mode rollout windows settle correctly.
- Gateway: ticket closes bill cost 0; the sweeper Reclaims Open tickets
after deliver_by WITHOUT closing live sessions (the conversation and its
affinity mirror survive their reclaimed ticket), gates reclaims on
deliver_by, and keeps an in-process settled-refs memory so terminal
indexer rows aren't retried every sweep forever.
Deployment (kimi-chat trio + gateway): ticket mode, 8 slots per supplier,
idle policy gateway 10 min / supplier 15 min (fixes the inversion where the
supplier idle-closed before the gateway janitor).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The demo endpoint lost its session on nearly every second message: affinity
required clients to re-echo the assistant reply byte-identically, so clients
that strip <think> blocks or reshape whitespace (Open WebUI, litellm) missed
every turn, opened a fresh escrow session on the next free single-slot
supplier, and left the old one pinned. Open WebUI's background title/tag-gen
requests each pinned another supplier — one message could exhaust all three
kimi replicas ("no available chat supplier").
- affinity: lenient assistant-content matching (think-strip + whitespace
collapse); user/system/tool messages and tool_calls stay byte-strict;
strict prefixes outrank lenient ones. The mirror is never rewritten with
client text, so receipt hashes stay valid.
- demoChat: trim trailing assistant/system messages (regenerate semantics)
instead of 400 "the last message must have role user or tool".
- supplier: roll the transcript back when the LLM call fails so retries
don't duplicate the delta (ChatSessionStore.truncateTranscript).
- demoChat: LRU-evict a demo session when no supplier is available (supplier
/v1/chat/end frees the slot, sweeper Accepts the escrow) and retry the
open with selectCandidates ignoreStatusFor to bypass the ~20s-stale
indexer status poll. Max 2 evictions per request.
- sweeper: idle janitor closes demo sessions unused for DEMO_SESSION_IDLE_MS
(default 3 min, below the supplier's 5 min CHAT_IDLE_TIMEOUT_MS) so
abandoned one-off sessions release their suppliers quickly.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
MVP demo: one operator-funded API key anyone can plug into an IDE/harness,
running real escrow-per-session marketplace conversations (llm.chat.v1)
behind the standard OpenAI /chat/completions wire shape.
- shared: ChatMessage gains tool role + tool_calls; normalizeChatMessage/
chatMessagesEquivalent keep supplier transcript and gateway mirror
hash-identical for receipts (field presence is part of the identity)
- supplier: /v1/chat/message accepts {messages,tools,tool_choice} alongside
legacy {content}; callOpenAiStream forwards tools, accumulates fragmented
delta.tool_calls, allows content-less tool turns; done frame carries the
assistant message verbatim + finish_reason + usage
- gateway: demo flag on api_keys (additive migration), create-demo-key CLI,
withdraw 403 + hidden deposit address for demo keys, per-IP demo rate
limit (demo keys skip the shared per-key limit), models filtered to chat
capability
- gateway executor (demoChat.ts): transcript-prefix affinity maps stateless
calls onto open sessions, forwards only unseen deltas, transparently
reopens on supplier idle-close; never falls back to one-shot (its settle
path assumes <=1 in-flight escrow per wallet)
- mutex scope: key mutex held only for chain work (open/close); off-chain
turns serialize on per-session locks so replicas serve users concurrently
- sweeper: chat-aware reclaim guard (35min vs 11min) and abandoned-session
cleanup; terminal state confirmed via on-chain queryUtxo (indexer keeps
lineage rows at their old state forever)
- deploy: kimi-chat replicas 2+3 compose files; chat idle timeout 2.5min
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The supplier sent a bare {model,messages,stream} to the OpenAI/OpenRouter
endpoint — no max_tokens and no reasoning control — so reasoning ("thinking")
models like kimi-k2.6 emitted reasoning tokens that could consume the
completion budget and finish with reason=length (truncated/empty answer =
failed job). deepseek-v4-flash has no thinking; kimi-k2.6 does by default.
Add two optional, env-gated request params (defaults preserve current behavior,
so ChatMock/DeepSeek-direct suppliers are unaffected):
- OPENAI_REASONING=off -> send reasoning:{enabled:false} (disable thinking:
faster, cheaper, straight answers; no reasoning-starved length truncation)
- OPENAI_MAX_TOKENS=N -> forward max_tokens:N. Must stay below the model
context (kimi context = 262144), so the advert's 262144 cap cannot be sent
verbatim (provider 400s when prompt+max_tokens > context).
openai.ts builds the body conditionally; config.ts parses the env; wired
through jobRunner.ts (text) and server.ts (chat-session stream).
Deployed on the 4 kimi + 3 deepseek-flash suppliers with OPENAI_REASONING=off
and OPENAI_MAX_TOKENS=131072 (env files are host-only/gitignored).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds a conversational supplier capability and a free demo, both streaming.
New paid `llm.chat.v1` type: one escrow per conversation (flat fee). "Start
chat" posts the escrow with a session-init placeholder prompt_hash and the
supplier Claims (reserving its single slot); turns then stream off-chain via
SSE with zero chain interaction; "End chat" has the supplier Submit a receipt
over the full transcript and the buyer Accept — the moment the user is charged.
The on-chain validator is content-agnostic (pins but never verifies the
hashes), so this reuses the existing contract unchanged. A supplier-side idle
watchdog auto-ends abandoned sessions before deliver_by.
Free Kimi K2.6 chat demo: a synthetic "demo:kimi-chat" tile that uses the same
chat UI but proxies straight to OpenRouter (no escrow, no payment), mirroring
the existing Piper demo.
Shared:
- postChatEscrow.ts + chatSessionPromptHash() (session-init prompt_hash helper
shared by the buyer builder and the supplier start gate).
Supplier:
- CAPABILITY_KIND="chat-session" (+ CHAT_IDLE_TIMEOUT_MS); callOpenAiStream()
streaming client; ChatSessionStore + endChatSession(); /v1/chat/{start,
message,end} handlers with idle + hard-cap watchdog timers.
Buyer:
- Marketplace.startChat()/endChat() with relaxed (session-init) receipt
verification + crypto verifyReceipt; /v1/chat/{start,message,end} routes
(SSE passthrough) and a free /v1/chat-demo/message OpenRouter proxy;
OPENROUTER_API_KEY/OPENROUTER_BASE_URL config; ChatForm.tsx + Dashboard
KIMI_DEMO_SUPPLIER tile and llm.chat.v1 dispatch.
Deploy:
- docker-compose.supplier-kimi-chat.yml + supplier/.env.kimi-chat.example.
Typecheck passes across all workspaces; full vitest suite green except 15
pre-existing time-bomb/RED failures (tx-claim/accept/post-advert live-CBOR
fixtures with hardcoded past deadlines + the M0-C golden-datum test), confirmed
identical on a clean tree.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Periodic setInterval ticker + on-failure backstop from the buildClaimTx
catch in server.ts. Both reuse the existing single-slot lock and call
runConsolidateWallet under the hood — planConsolidate short-circuits to
already-healthy in the steady state so no tx, no fee on a clean wallet.
Driven by deepseek-lane outage 2026-05-28 (supplier wallet drifted to 14
UTxOs with 9 dust 2-AP3X outputs; Claim tx coin-selector picked dust as
working input → chain_submit_failed; lane stalled until manual
tx:consolidate-wallet).
buyer/scripts/monitor-wallets: surface deepseek + kimi as operator wallets.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds an optional OPENAI_API_KEY env var that supplier/src/openai.ts forwards as
`Authorization: Bearer <key>`. Empty/unset preserves the ChatMock behaviour
(no auth header). Unlocks pointing the openai backend at DeepSeek directly
(https://api.deepseek.com) — no local proxy needed.
New compose: deploy/mainnet/docker-compose.supplier-deepseek.yml (third mainnet
supplier, model deepseek-v4-pro, mp-suppliers-deepseek.vector.apexfusion.org).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a one-shot operational tx that re-shapes any operator wallet to
{5 AP3X collateral candidate, working remainder} so subsequent Plutus
script-spends (Claim/Submit/Accept/Reclaim/RetireAdvert) find a UTxO
that satisfies lucid's setCollateral=5_000_000n default coin selector.
Symptom this addresses: open-webui mainnet supplier wallet held 8 UTxOs
max 2.40 AP3X, every Accept tx failed at lucid .complete() with the
"wallet must hold a pure-ADA UTxO ≥ 5 ADA" error path. Same fragmentation
risk on the other 4 mainnet operator wallets (one big UTxO surrounded by
dust — first script-spend consumes the big one and leaves nothing usable
as collateral).
Phase 1 — new tx:consolidate-wallet CLI (supplier + buyer):
- planConsolidate.ts — pure planner; supports consolidate (N>1 → 2),
split (1 → 2), and already-healthy skip.
- consolidateWallet.ts — live builder; presetWalletInputs pins all
UTxOs, two pay.ToAddress outputs (collateral + working).
- runConsolidateWallet.ts — orchestration (build + submit + awaitTx).
- supplier/src/cli/consolidate-wallet.ts + buyer mirror.
- monitor-wallets.ts — add supplier-tts@vector-marketplace (was
misclassified as 3rd-party).
Phase 2 — defensive pre-check in the 5 script-spend builders:
- assertCollateralCandidate(utxos) fires at the top of
buildSpendOpenOrClaimedTx (Claim/Submit), buildLiveTxForAccept,
buildLiveTxForReclaim, buildLiveTxForRetireAdvert. Throws a
TxConstructionError("collateral_required") pointing at
tx:consolidate-wallet instead of waiting for lucid's deep failure.
Tests: 12 new passing (planConsolidate edge cases + assert helper).
Pre-existing lucid 0.4.30 WASM "unreachable" failures in tx-*-live.test.ts
fixtures are unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- jobRunner.ts: console.warn at each jobs.fail site (chat + tts paths, 6 sites). Three mainnet GPT-supplier 502s on 2026-05-19 left no trace beyond Traefik — adding a one-line log restores grep-ability.
- docker-compose.supplier-gpt.yml: OPENAI_TIMEOUT_MS 180000→300000. The 3 observed 502s were all exactly 90 polls × 2s, i.e. the supplier's openai_timeout firing.
- docker-compose.chatmock.yml: serve --reasoning-effort low --reasoning-summary none. Default 'medium'/'auto' pushed GPT-5.4 think-time over the prior 180s budget.
Note: on-chain advert ADVERT_MAX_PROCESSING_MS still pins 180000 until a re-advert tx is posted — buyers compare against the on-chain value, not the runtime env.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mirrors deploy/testnet/docker-compose.supplier-tts.yml and the simulation
PiperTTS stack, co-located on vector-marketplace as a private internal
service on marketplace-mainnet-net (no Traefik labels, no public DNS for
Piper itself). Supplier reaches Piper at http://marketplace-mainnet-tts-piper:8000.
Voice: en_US-lessac-high. New supplier wallet + on-chain advert are
published out-of-band by the operator per the runbook header.
Files:
- deploy/mainnet/docker-compose.tts-piper.yml openedai-speech-min, private
- deploy/mainnet/docker-compose.supplier-tts.yml CAPABILITY_KIND=tts supplier
- deploy/mainnet/install-voice-extras.sh voice download + yaml mapping
- deploy/mainnet/voice-mappings.py idempotent yaml editor
- supplier/.env.tts.example operator env template
- .gitignore whitelist .env.*.example
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a Codex/ChatGPT backend path for the supplier alongside the
existing Ollama one. The mainnet supplier is repointed at a localhost
ChatMock proxy (RayBytes/ChatMock, pinned :v1.35) that authenticates
against a ChatGPT Pro subscription via OAuth.
- supplier/src/openai.ts: HTTP client for OpenAI-compatible
/v1/chat/completions. Same return shape as ollama.callOllama so
jobRunner can branch cleanly.
- supplier/src/jobRunner.ts: branch on config.llmBackend.
- supplier/src/config.ts: LLM_BACKEND ("ollama" default | "openai"),
OPENAI_BASE_URL, OPENAI_TIMEOUT_MS (default 180_000 — GPT first-token
latency).
- supplier/src/cli/gen-keypair.ts: standalone Ed25519 keypair generator
+ bech32 address derivation. Used to bootstrap a fresh supplier
wallet without round-tripping through tx:post-advert --dry-run.
- deploy/mainnet/docker-compose.chatmock.yml: long-running chatmock
serve + one-shot --profile login flow.
- deploy/mainnet/docker-compose.supplier.yml: rewritten for openai
backend; Traefik routes at mp-suppliers-gpt.vector.apexfusion.org so
the unsuffixed slot stays free for a future supplier with a
different backend.
- tests/unit/supplier-openai.test.ts: unit coverage for callOpenAi.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Splits the single `lovelacePerOutput` param in publishReferenceScripts.ts
into `escrowLovelace` and `advertLovelace`, exposed via env vars
ESCROW_PUBLISH_LOVELACE / ADVERT_PUBLISH_LOVELACE in the CLI. Defaults
unchanged at 30 AP3X each (60 AP3X total).
Motivation: mainnet rollout where the publishing wallet has only ~57 AP3X
(buyer wallet) — needs to publish at calculated min-utxo (~17 AP3X escrow
+ ~7 AP3X advert = 24 AP3X total) instead of the safe 60 AP3X default.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Static re-exports from tx/index.ts caused vite to trace
loadEscrowScript/loadAdvertScript/buildLiveTxForPublishReferenceScripts
into the buyer UI bundle, hitting the CML WASM "ESM integration proposal"
limitation. Sibling builders avoid this with /* @vite-ignore */ dynamic
imports; we keep symmetry by parking the helpers on a separate
@marketplace/shared/tx/server entry that the buyer never imports.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When ESCROW_REF_UTXO / ADVERT_REF_UTXO env vars are set (format
<txhash>#<idx>), Claim/Submit/Accept/Reclaim/RetireAdvert call
.readFrom([refUtxo]) instead of inlining the validator CBOR via
.attach.SpendingValidator(). Lucid-evolution 0.4.30 auto-extracts
utxo.scriptRef into the script-witness map (dist/index.js:1903–1913),
so no separate attach call is needed. Default (env unset) is unchanged.
Adds buildLiveTxForPublishReferenceScripts (a one-shot tx that
publishes both validators as scriptRef-bearing UTxOs at a caller-
supplied address — typically the all-zero-pkh enterprise burn address)
and the supplier tx:publish-reference-scripts CLI that wraps it.
Targeted at testnet trial; mainnet rollout is a separate flip after
verification.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes 2/4 builders from the M1-F-D-defer list (postAdvert already live in
7ea7328; updateAdvert and release remain).
Reclaim:
- buildLiveTxForReclaim in liveCbor.ts (mirrors buildLiveTxForAccept):
spends the escrow UTxO with REDEEMER_RECLAIM (Constr3), pays
payment + buyer_bond + supplier_bond to buyer, delegates UPLC eval
to Ogmios via localUPLCEval:false. Validity range = tipMs ± 60s.
- reclaim.ts: replace the unconditional mockSlotToWallclockMs tipMs
computation with `isLive ? Date.now() : mockSlotToWallclockMs(tipSlot)`,
matching accept.ts:77. Adds the live-CBOR dispatch branch.
Funds-at-risk before this fix: with the broken tipMs path, every live
Reclaim was rejected client-side ("tip < deliver_by" using a Nov-1970
tip). If a supplier never Claimed, buyer's 6-7 AP3X stayed locked
indefinitely. Verified recovery of 4 stuck escrows on Vector L2 testnet
(txs f6660e52, 769f16fd, 79970d86, 9f4db14b — 14 AP3X total recovered).
RetireAdvert:
- buildLiveTxForRetireAdvert in liveCbor.ts: spends advert UTxO with
REDEEMER_RETIRE_ADVERT (Constr2), advert script address.
- retireAdvert.ts: live-CBOR dispatch.
- supplier CLI: retire-advert.ts + retireAdvertFlow.ts.
- supplier/package.json: tx:retire-advert script.
Indexer:
- follower/blockProcessor + worker: handle RetireAdvert events,
transition advert row state to Retired.
- db/cache.ts: markAdvertTerminal helper.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
