fix(mcp): real quota display (% + $), billing-aware unlimited, quota-not-credits enrichment#152
Merged
Merged
Conversation
…imited billing-aware leadbay_account_status described quota in raw 'credits'. Now it mirrors the web app: a % used + $ spend gauge per daily/weekly/monthly window (from quota_status spend[]), with a per-resource breakdown fallback when no spend cap exists. Quota is included on any account-status question when readable. Also fixes the unlimited-account rule: an @leadbay.ai account is unlimited ONLY when billing is disabled (backend seat sentinel 100000, or billing absent). @leadbay.ai with billing ENABLED (metered freemium/paid) is now treated as a real user — real quota gauge, real credit gate. Co-Authored-By: Claude <noreply@anthropic.com>
After a quota-consuming action completes (bulk enrich all_done, enrich_contacts) or a confirmed top-up, the agent now calls leadbay_account_status and shows the per-window %/$ quota gauge — the canonical surface — instead of only a bare credits line. Skipped for unlimited accounts, unreadable quota, or no billing. Co-Authored-By: Claude <noreply@anthropic.com>
…channel
The client pre-refused enrichment when billing.ai_credits <= 0, but ai_credits
is credits CONSUMED (an accumulator starting at 0) — so freemium/fresh accounts
with quota still available were falsely blocked ('No enrichment credits
remaining'). Removed the client-side gate; the backend 429 is the real quota
gate. Reframed enrich prompts to talk QUOTA (email + phone reveals consume
quota), dropping the confusing 'credits' framing and the fabricated 'You have N
credits' line. When the user asks to enrich without naming a channel, the agent
now asks via ask_user_input_v0 whether to add phone (email is the default).
Superseded the obsolete credit-gate tests (enrich-contacts.test.ts rewritten;
enrich-contacts-unlimited.test.ts → enrich-contacts-quota-gate.test.ts).
Co-Authored-By: Claude <noreply@anthropic.com>
The runtime elicitation prompt (shown directly by the host on an elicitation-capable path) said 'This spends credits (balance: N)' — bypassing the template reframe. Rewrote it to talk quota (email/phone reveals consume quota) and reframed the credits_remaining output-schema description to advisory-only (never a spend gate). Addresses Codex review on #152. Co-Authored-By: Claude <noreply@anthropic.com>
…quota-understanding-in-the-mcp # Conflicts: # packages/core/src/tool-descriptions.generated.ts # packages/promptforge/tool-descriptions/composite/bulk-enrich-status.md.tmpl
ArtyETH06
changed the title
…NS_EXTRA_REFILL The new quota guidance preferred quota.user and referred to the resource as lowercase lens_extra_refill. But leadbay_extend_lens pre-checks are ORG-scoped (extend-lens.ts reads quota.org.resources) and match the uppercase LENS_EXTRA_REFILL. Carved out the pre-check: user preference stays for the display gauge only; the LENS_EXTRA_REFILL pre-check reads quota.org.resources[] and matches case-insensitively (both casings). Addresses Codex review on #152. Co-Authored-By: Claude <noreply@anthropic.com>
…untime lookup readExtraRefillQuota() matched the resource type with exact-case === "LENS_EXTRA_REFILL", but the live quota API emits lowercase lens_extra_refill — so a 429 from leadbay_extend_lens nulled out used_today / resets_at even though the backend supplied the row. Now normalizes case (toUpperCase compare), matching the lowercase shape this PR documents as valid. New test file covers the lowercase + mixed-case rows. Addresses Codex review on #152. Co-Authored-By: Claude <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e4446b4486
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
1. Preserve unlimited when seats absent: @leadbay.ai + billing present but no seats is now treated as unknown/legacy-disabled (unlimited), not metered — fixes a product#3851 regression for the pre-3865 internal shape. 2. Granular enrich: credits_remaining is advisory/non-displayable — never print _(N credits remaining)_ (0 falsely reads as out-of-credits on quota-gated). 3. Post-spend refresh: leadbay_enrich_contacts only LAUNCHES async work, so defer the quota refresh to bulk all_done / get_contacts enrichment.done — not immediately after the granular launch. 4. Keep tests in new files: restored enrich-contacts.test.ts to main minus the obsolete throw test; new quota-gate coverage lives in the new file. Co-Authored-By: Claude <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0e87a0e3ea
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
1. bulk-enrich-status: credits_remaining is advisory/non-displayable — stop allowing _(N credits remaining)_ (a consumed counter reads 0 on a quota-backed account, falsely 'out of credits'). account_status gauge is the only quota surface. 2. enrich-titles no-channel elicitation now offers a phone opt-in (include_phone toggle) instead of silently confirming email-only, honoring the promised 'email vs email + phone' choice. New test file covers it. 3. lens-refill pre-check falls back to quota.user.resources[] when quota.org is absent (non-admin callers only get the user group) — else the row is invisible and the write burns the 429 the pre-check exists to avoid. Applied in the snippet, template, account-status schema, and extend-lens runtime. Co-Authored-By: Claude <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5b8adb6406
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
1. bulk-enrich-status OUTPUT SCHEMA: credits_remaining reframed advisory/ non-displayable (the template was fixed earlier; the schema still called it a balance — MCP clients see both). 2. Direct extend-lens prompts (extend-lens.md.tmpl + leadbay_extend_my_lens): lens-refill pre-check now org-first then user-fallback + case-insensitive, matching the account-status guidance. 3. WORKFLOWS.md 34 & 42: drop the 'surface credits_remaining' requirement — surface enrichable_contacts only; enrichment consumes quota, credits not displayed. Keeps the normative contract aligned with the prompt change. 4. get-quota template/description: prefer user group (org admin-only), describe spend[] %/$ gauge + resources[] + topup, case-insensitive types, no 'credits'. 5. account-status rendering_hint: prefer quota.user, fall back to quota.org when user absent (admins/legacy org-only shapes keep their display). Co-Authored-By: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What & why
The MCP talked in raw "credits" for quota, and a freemium user with quota left was wrongly told they couldn't enrich. This unifies the full fix for #3865 (folds in the enrichment work from the former #153).
1. Quota renders like the frontend.
leadbay_account_statusshows a per-window % used + $ spend gauge (daily/weekly/monthly, fromquota_statusspend[]), with a per-resource breakdown fallback when the plan has no spend cap. Included on any account question (incl. "what account am I connected to?"); silent only when unreadable, unlimited, or null.2. Billing-aware unlimited.
@leadbay.aiis unlimited only when billing is disabled (seat sentinel100000, or billing absent). Metered freemium/paid@leadbay.aiis treated as a real user — real gauge, real gate.3. Enrichment is gated by quota, not credits. Removed the buggy client-side
ai_credits <= 0refusal (ai_creditsis credits CONSUMED, starts at 0 — it falsely blocked freemium accounts). The backend 429 is the real gate. Reframed all enrich prompts + the runtime elicitation copy from "credits" to quota (email + phone reveals consume quota). "Enrich" with no channel → the agent asks email-only vs email+phone viaask_user_input_v0.4. Refreshed quota after a paid action. After a completed enrichment or a confirmed top-up, the agent shows the refreshed per-window gauge.
Changes
rendering/quota-windows.md(new) +account-status.md.tmpl— the gauge render._credits-helpers.tsbilling-awareisUnlimitedAccount;types.ts—billing.seats,QuotaResourceUsage.max_units,topup, nullableplan/org.tools/enrich-contacts.ts— dropped the credit gate; advisory read only.enrich-titles.tsruntime elicitation — quota-not-credits + ask-about-phone.server-instructions/quota-topup.md+bulk-enrich-status.md.tmpl— show-quota-after-spend.account-status-quota-render.test.ts,unlimited-billing-aware.test.ts,enrich-contacts-quota-gate.test.ts. Superseded the obsolete credit-gate contract by removingunlimited-account.test.ts+enrich-contacts-unlimited.test.tsand rewritingenrich-contacts.test.ts(delete+rewrite, per repo owner's explicit direction — the obsolete assertions tested the exact bug being fixed; all still-valid coverage carried over).Verification
pnpm -r build && pnpm -r test && pnpm -r typecheck— green (core 539, mcp 525, promptforge 16, components 47).@leadbay.aifreemium now returnsunlimited_credits: false+ populatedspend[]; the gauge renders; enrichment no longer false-refuses.No eval run (author-run; prompt-behaviour change). WORKFLOW 30 (silent-on-unreadable-quota) stays green.
Closes https://github.com/leadbay/product/issues/3865