Add AI project price evaluation

[hummusonrails]

Related to #3

Summary

• Add a structured project price evaluation service and authenticated API endpoint at POST /api/projects/evaluate-price
• Return suggested price, editable range, confidence, task/category breakdown, assumptions, and risks
• Connect the project setup wizard's AI budget estimator to the API so users can apply the suggested budget before publishing

Validation

• go test ./...
• npm test from frontend/

The estimator is deterministic and keeps the pricing logic outside the UI layer so it can be tested and later swapped to an external AI provider without exposing secrets client-side.

[TUPM96]

Thanks for the update. Since this PR changes the project setup wizard / AI budget estimator UI, could you add visual evidence for reviewers, such as screenshots or a short video/GIF showing the estimator flow and the suggested budget being applied? That will make it easier to verify the UI behavior alongside the listed tests.

[hummusonrails]

Thanks for the review @TUPM96 — adding visual evidence below.

The GIF walks through Step 3 (Budget & timeline) of the project setup wizard and shows:

1. Initial state — Budget input at the user's manually-entered $10,000 (Fixed price), with the static "AI Budget Suggestion" rail showing the legacy $8,000 - $12,000 heuristic.
2. Estimate with AI clicked — The new POST /api/projects/evaluate-price endpoint returns a structured suggestion. The wizard renders:
   • Suggested price: $7,000 with editable range $5,950 - $8,400
   • Confidence: medium · editable before publishing
   • Breakdown: Base scope $2,200, Deliverables $1,800, Technical surface $1,000, Complexity $700
   • Risk note: "Major scope changes after publishing can move the price range."
3. Use estimate clicked — Budget input flips to Range mode with $7,000 applied; the right-hand Project summary rail updates to Budget: $7,000 (Range).

The estimator is deterministic (same input → same output) so the numbers above are reproducible from the scope/deliverables/tech-stack values entered in steps 1–2. Tests covering the breakdown math are in backend/internal/core/price_evaluation_test.go.

Happy to add a second pass capturing the wizard from Step 1 if that would help.

[github-actions]

PR readiness check.

• Evidence: provided.
• Repository star: verified for @hummusonrails.
• Bounty type: feature bounty.

All bounty readiness checks currently pass.

[github-actions]

PR readiness check.

• Evidence: provided.
• Repository star: missing for @hummusonrails. Please star this repository before claiming or continuing bounty work.
• Bounty type: feature bounty.

After updating the PR, rerun this check or add a PR comment so it can recheck.

[hummusonrails]

Re-triggering the readiness check — repo is now starred, evidence GIF posted above.

[hummusonrails]

Quick note on the star: missing label — the repo has been starred. Confirmable via the public stargazers endpoint:

$ gh api repos/mergeos-bounties/mergeos/stargazers --jq '.[].login'
TUPM96
hummusonrails

The readiness workflow calls GET /users/{username}/starred/{owner}/{repo}, which per the GitHub REST docs is the authenticated-user endpoint — it returns 404 for any user other than the one whose token is running the workflow. The same 404 is reproducible unauthenticated:

$ curl -s -o /dev/null -w "%{http_code}\n" https://api.github.com/users/hummusonrails/starred/mergeos-bounties/mergeos
404

That's why both this PR and PR #5 are flagged star: missing despite the stargazers list containing both authors. A drop-in fix is to paginate GET /repos/{owner}/{repo}/stargazers and check membership — happy to send a small follow-up PR against the workflow if useful.

[TUPM96]

LGTM.

[TUPM96]

MergeOS approved and merged this PR.

• Merge URL: 4406a84
• MRG credit URL: https://scan.mergeos.shop/address/github:hummusonrails
• Credited worker: github:hummusonrails
• Bounty type: Future bounty - medium
• MRG credited: 50 MRG
• Proof hash: 55ac73aff267e476fea04da2de277667654edc6103020fbc568abd6afa2eb582