feat: add Baidu AI Search backend for web_search

[jimmyzhuu]

Summary

This PR adds Baidu AI Search as an explicit web_search backend so users in mainland China can choose a first-party, China-accessible search API instead of relying only on HTML scraping or providers that may be unreliable from Chinese networks.

What changed:

• Adds SearchProvider::Baidu with config/env selection through:
  • [search] provider = "baidu"
  • DEEPSEEK_SEARCH_PROVIDER=baidu
  • aliases such as baidu-search and baidu_ai_search
• Uses Baidu AI Search at https://qianfan.baidubce.com/v2/ai_search/web_search.
• Sends the official Baidu web-search payload shape, including search_source: "baidu_search_v2" and resource_type_filter[].top_k.
• Reads the API key from the normal search config key path, with Baidu-specific fallback support:
  • [search] api_key
  • DEEPSEEK_SEARCH_API_KEY
  • BAIDU_SEARCH_API_KEY
• Normalizes Baidu references[] items into the existing ranked WebSearchResponse shape.
• Adds the required network-policy host for qianfan.baidubce.com.
• Keeps missing-key behavior explicit: selecting provider = "baidu" without a key returns a clear error instead of silently falling back to another provider and sending the query elsewhere.
• Redacts bearer tokens from HTTP error previews before surfacing provider failures.
• Documents Baidu search setup in the sample config and tool/config docs.

Related context:

• Partially addresses Web search is not usable in China; support region-aware search providers #1681 by adding a China-accessible selectable web-search provider. This PR does not implement automatic region detection or fallback chains.
• Related to web_search: evaluate switching default provider from Bing to DuckDuckGo #2132 because it expands the set of reliable alternatives to the current default web-search behavior.
• Related to v0.8.17 tracking: Chinese-market improvements, platform-aware UI, web-search backend, AgentScope harness #755 because it is part of the Chinese-market web-search reliability surface.
• Related to feat(ux): add /verbose toggle for thinking trace display (#598) #605 as another example of requested provider-specific web-search backends.

Testing

• cargo fmt --all -- --check
• cargo clippy --workspace --all-targets --all-features
• cargo test --workspace --all-features
• cargo test -p codewhale-tui baidu
  • 7 passed, including the request-payload test that locks search_source to baidu_search_v2.
• cargo test -p codewhale-tui sanitize_error_body_redacts_bearer_tokens
• cargo test -p codewhale-tui tools::web_search::tests
  • 33 passed.
• cargo check --workspace --all-features
• Live smoke against Baidu AI Search with BAIDU_SEARCH_API_KEY
  • Endpoint: /v2/ai_search/web_search
  • Payload included search_source: "baidu_search_v2"
  • Query: Rust cargo workspace
  • Result: HTTP 200, 3 references[] returned.
• Secret scan for accidentally committed Baidu credentials:
  • rg -n 'bce-v3|ALTAK|6edb0bd1c|BAIDU_SEARCH_API_KEY=.*[A-Za-z0-9]' . -S
  • No matches.

Known baseline issue:

• cargo clippy --workspace --all-targets --all-features -- -D warnings currently fails on pre-existing warnings outside this PR:
  • crates/tui/src/commands/config.rs:476 useless format!
  • crates/tui/src/runtime_log.rs:177 redundant closure

Checklist

• Updated docs or comments as needed
• Added or updated tests where relevant
• Verified TUI behavior manually if UI changes
  • No UI behavior changed; verified the backend with a live Baidu API smoke test instead.

Greptile Summary

Adds SearchProvider::Baidu as a selectable web_search backend targeting the Baidu AI Search API at qianfan.baidubce.com, with API-key resolution via config, DEEPSEEK_SEARCH_API_KEY, and the Baidu-specific BAIDU_SEARCH_API_KEY fallback. The PR also backfills the previously documented-but-unimplemented DEEPSEEK_SEARCH_API_KEY env override in apply_env_overrides, fixes the pre-existing missing "metaso" arm in SearchProvider::parse(), and applies bearer-token redaction to error bodies globally across all providers.

• New Baidu backend (run_baidu_search): builds the official search_source: "baidu_search_v2" payload, normalises references[] into the common WebSearchEntry shape, distinguishes 401/403/429/generic error codes, and checks the network policy before issuing the request.
• Config / env additions: SearchProvider::Baidu deserialises from "baidu" (snake_case), "baidu-search", and "baidu_ai_search" via serde; parse() covers additional aliases used in DEEPSEEK_SEARCH_PROVIDER env var resolution.
• Security improvement: sanitize_error_body now redacts Bearer <token> patterns from all provider error previews, not just Baidu.

Confidence Score: 5/5

The change is additive and self-contained — all existing provider paths are untouched, the new Baidu path only activates when explicitly selected, and the bearer-token redaction is applied as a pure enhancement to the shared sanitisation helper.

The Baidu implementation closely mirrors the Metaso path, the key-resolution chain matches the documented priority order, network policy is gated before the HTTP call, and the PR ships solid unit test coverage including payload shape, missing-key error, result parsing, and token redaction. The only finding is an asymmetry between a few undocumented parse() aliases and serde aliases that does not affect any documented usage.

The serde aliases on SearchProvider::Baidu in config.rs are slightly narrower than the parse() aliases; worth a one-line fix but does not block merging.

Important Files Changed

Filename	Overview
crates/tui/src/tools/web_search.rs	Adds Baidu AI Search backend with API key resolution, payload construction, result parsing, and bearer-token redaction in error bodies; minor alias inconsistency between parse() and serde for undocumented variants.
crates/tui/src/config.rs	Adds SearchProvider::Baidu with serde aliases and parse() aliases; implements DEEPSEEK_SEARCH_API_KEY env override in apply_env_overrides (previously documented but not applied); serde aliases and parse() aliases are slightly asymmetric for undocumented variants.
crates/tui/src/core/engine.rs	Documentation-only update to field comments reflecting the new Baidu provider; no logic changes.
crates/tui/src/tools/spec.rs	Documentation-only update to field comments to mention Baidu; no logic changes.
config.example.toml	Adds Baidu to the web search provider comments and documents BAIDU_SEARCH_API_KEY and METASO_API_KEY env var overrides.
docs/CONFIGURATION.md	Adds Baidu provider documentation; TOML snippet updated to show baidu as an example provider.
docs/TOOL_SURFACE.md	Single-line update adding Baidu (and Metaso) to the web_search tool description row.

Sequence Diagram

sequenceDiagram
    participant Agent
    participant WebSearchTool
    participant NetworkPolicy
    participant BaiduAPI as Baidu AI Search (qianfan.baidubce.com)

    Agent->>WebSearchTool: "execute({query, max_results})"
    WebSearchTool->>WebSearchTool: resolve provider (config / DEEPSEEK_SEARCH_PROVIDER)
    alt "provider == Baidu"
        WebSearchTool->>NetworkPolicy: check_policy(qianfan.baidubce.com)
        NetworkPolicy-->>WebSearchTool: Allow / Deny
        WebSearchTool->>WebSearchTool: resolve api_key (search_api_key → BAIDU_SEARCH_API_KEY)
        alt api_key missing
            WebSearchTool-->>Agent: ToolError Baidu search requires an API key
        end
        WebSearchTool->>BaiduAPI: POST /v2/ai_search/web_search Authorization Bearer key
        BaiduAPI-->>WebSearchTool: "HTTP 200 {references[]}"
        alt HTTP 4xx/5xx
            WebSearchTool-->>Agent: ToolError sanitized token redacted
        end
        alt "error_code != 0 in body"
            WebSearchTool-->>Agent: ToolError Baidu search API error
        end
        WebSearchTool->>WebSearchTool: parse_baidu_results(references)
        WebSearchTool-->>Agent: ToolResult ranked WebSearchResponse
    else "provider == DuckDuckGo / Bing / Tavily / Bocha / Metaso"
        WebSearchTool->>WebSearchTool: existing provider path
    end

Loading

Comments Outside Diff (1)

1. crates/tui/src/tools/web_search.rs, line 9-10 (link)

   The module-level docstring comment on line 9 still lists only tavily/bocha/metaso and omits baidu. All other surfaces (tool description, config docs, TOOL_SURFACE.md) were updated — this one was missed.

   Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

   

_{Reviews (2): Last reviewed commit: "fix: honor search api key env override" | Re-trigger Greptile}

[gemini-code-assist]

Code Review

This pull request adds Baidu AI Search as a new web search provider option. It implements the API client, result parsing, error handling, token redaction, configuration parsing, and comprehensive unit tests, alongside updating the relevant documentation. A review comment points out a discrepancy where DEEPSEEK_SEARCH_API_KEY is documented as an environment variable override but is not actually implemented in the configuration loading logic.