feat: add citation/reference support to deep research reports (#1143)

* feat: add citation/reference support to deep research reports (#1141)

- Enhance lead agent system prompt with mandatory citation requirements
  after web_search/web_fetch tool usage
- Add citation examples and best practices to GitHub Deep Research skill
- Add citation hints to report template (Executive Summary, Key Analysis)
- Style regular markdown links in frontend for visual distinction
  (color, underline, hover effect)
- Fix TitleMiddleware being registered when title generation is disabled

* fix: address PR review comments

- Revert TitleMiddleware conditional registration (agent.py) to avoid
  sync/async incompatibility with DeerFlowClient
- Fix markdown link rendering: merge classNames instead of overwriting,
  only set target=_blank for external http(s) URLs
- Remove unrelated package.json/pnpm-lock.yaml changes

* fix: use plain markdown links in Sources section for cleaner rendering

Inline citations in report body use [citation:Title](URL) for pill/badge style.
Sources section uses plain [Title](URL) for simple underlined link style.

* fix(frontend): render plain links as underlined text in artifact markdown

Only links with citation: prefix render as Badge pills.
Regular links in Sources section now render as underlined text links.

---------

Co-authored-by: Willem Jiang <willem.jiang@gmail.com>

backend/packages/harness/deerflow/agents/lead_agent/prompt.py

@@ -257,15 +257,66 @@ You: "Deploying to staging..." [proceed]
 </response_style>
 
 <citations>
-- When to Use: After web_search, include citations if applicable
+**CRITICAL: Always include citations when using web search results**
-- Format: Use Markdown link format `[citation:TITLE](URL)`
+
-- Example: 
+- **When to Use**: MANDATORY after web_search, web_fetch, or any external information source
+- **Format**: Use Markdown link format `[citation:TITLE](URL)` immediately after the claim
+- **Placement**: Inline citations should appear right after the sentence or claim they support
+- **Sources Section**: Also collect all citations in a "Sources" section at the end of reports
+
+**Example - Inline Citations:**
 ```markdown
 The key AI trends for 2026 include enhanced reasoning capabilities and multimodal integration
 [citation:AI Trends 2026](https://techcrunch.com/ai-trends).
 Recent breakthroughs in language models have also accelerated progress
 [citation:OpenAI Research](https://openai.com/research).
 ```
+
+**Example - Deep Research Report with Citations:**
+```markdown
+## Executive Summary
+
+DeerFlow is an open-source AI agent framework that gained significant traction in early 2026
+[citation:GitHub Repository](https://github.com/bytedance/deer-flow). The project focuses on
+providing a production-ready agent system with sandbox execution and memory management
+[citation:DeerFlow Documentation](https://deer-flow.dev/docs).
+
+## Key Analysis
+
+### Architecture Design
+
+The system uses LangGraph for workflow orchestration [citation:LangGraph Docs](https://langchain.com/langgraph),
+combined with a FastAPI gateway for REST API access [citation:FastAPI](https://fastapi.tiangolo.com).
+
+## Sources
+
+### Primary Sources
+- [GitHub Repository](https://github.com/bytedance/deer-flow) - Official source code and documentation
+- [DeerFlow Documentation](https://deer-flow.dev/docs) - Technical specifications
+
+### Media Coverage
+- [AI Trends 2026](https://techcrunch.com/ai-trends) - Industry analysis
+```
+
+**CRITICAL: Sources section format:**
+- Every item in the Sources section MUST be a clickable markdown link with URL
+- Use standard markdown link `[Title](URL) - Description` format (NOT `[citation:...]` format)
+- The `[citation:Title](URL)` format is ONLY for inline citations within the report body
+- ❌ WRONG: `GitHub 仓库 - 官方源代码和文档` (no URL!)
+- ❌ WRONG in Sources: `[citation:GitHub Repository](url)` (citation prefix is for inline only!)
+- ✅ RIGHT in Sources: `[GitHub Repository](https://github.com/bytedance/deer-flow) - 官方源代码和文档`
+
+**WORKFLOW for Research Tasks:**
+1. Use web_search to find sources → Extract {{title, url, snippet}} from results
+2. Write content with inline citations: `claim [citation:Title](url)`
+3. Collect all citations in a "Sources" section at the end
+4. NEVER write claims without citations when sources are available
+
+**CRITICAL RULES:**
+- ❌ DO NOT write research content without citations
+- ❌ DO NOT forget to extract URLs from search results
+- ✅ ALWAYS add `[citation:Title](URL)` after claims from external sources
+- ✅ ALWAYS include a "Sources" section listing all references
 </citations>
 
 <critical_reminders>

frontend/src/components/workspace/artifacts/artifact-file-detail.tsx

@@ -38,7 +38,7 @@ import { checkCodeFile, getFileName } from "@/core/utils/files";
 import { env } from "@/env";
 import { cn } from "@/lib/utils";
 
-import { CitationLink } from "../citations/citation-link";
+import { ArtifactLink } from "../citations/artifact-link";
 import { useThread } from "../messages/context";
 import { Tooltip } from "../tooltip";
 
@@ -274,7 +274,7 @@ export function ArtifactFilePreview({
         <Streamdown
           className="size-full"
           {...streamdownPlugins}
-          components={{ a: CitationLink }}
+          components={{ a: ArtifactLink }}
         >
           {content ?? ""}
         </Streamdown>

frontend/src/components/workspace/citations/artifact-link.tsx

@@ -0,0 +1,33 @@
+import type { AnchorHTMLAttributes } from "react";
+
+import { cn } from "@/lib/utils";
+
+import { CitationLink } from "./citation-link";
+
+function isExternalUrl(href: string | undefined): boolean {
+  return !!href && /^https?:\/\//.test(href);
+}
+
+/** Link renderer for artifact markdown: citation: prefix → CitationLink, otherwise underlined text. */
+export function ArtifactLink(props: AnchorHTMLAttributes<HTMLAnchorElement>) {
+  if (typeof props.children === "string") {
+    const match = /^citation:(.+)$/.exec(props.children);
+    if (match) {
+      const [, text] = match;
+      return <CitationLink {...props}>{text}</CitationLink>;
+    }
+  }
+  const { className, target, rel, ...rest } = props;
+  const external = isExternalUrl(props.href);
+  return (
+    <a
+      {...rest}
+      className={cn(
+        "text-primary underline decoration-primary/30 underline-offset-2 hover:decoration-primary/60 transition-colors",
+        className,
+      )}
+      target={target ?? (external ? "_blank" : undefined)}
+      rel={rel ?? (external ? "noopener noreferrer" : undefined)}
+    />
+  );
+}

frontend/src/components/workspace/messages/markdown-content.tsx

@@ -1,16 +1,21 @@
 "use client";
 
 import { useMemo } from "react";
-import type { HTMLAttributes } from "react";
+import type { AnchorHTMLAttributes } from "react";
 
 import {
   MessageResponse,
   type MessageResponseProps,
 } from "@/components/ai-elements/message";
 import { streamdownPlugins } from "@/core/streamdown";
+import { cn } from "@/lib/utils";
 
 import { CitationLink } from "../citations/citation-link";
 
+function isExternalUrl(href: string | undefined): boolean {
+  return !!href && /^https?:\/\//.test(href);
+}
+
 export type MarkdownContentProps = {
   content: string;
   isLoading: boolean;
@@ -30,7 +35,7 @@ export function MarkdownContent({
 }: MarkdownContentProps) {
   const components = useMemo(() => {
     return {
-      a: (props: HTMLAttributes<HTMLAnchorElement>) => {
+      a: (props: AnchorHTMLAttributes<HTMLAnchorElement>) => {
         if (typeof props.children === "string") {
           const match = /^citation:(.+)$/.exec(props.children);
           if (match) {
@@ -38,7 +43,16 @@ export function MarkdownContent({
             return <CitationLink {...props}>{text}</CitationLink>;
           }
         }
-        return <a {...props} />;
+        const { className, target, rel, ...rest } = props;
+        const external = isExternalUrl(props.href);
+        return (
+          <a
+            {...rest}
+            className={cn("text-primary underline decoration-primary/30 underline-offset-2 hover:decoration-primary/60 transition-colors", className)}
+            target={target ?? (external ? "_blank" : undefined)}
+            rel={rel ?? (external ? "noopener noreferrer" : undefined)}
+          />
+        );
       },
       ...componentsFromProps,
     };

skills/public/github-deep-research/SKILL.md

@@ -147,5 +147,20 @@ Save report as: `research_{topic}_{YYYYMMDD}.md`
 3. **Triangulate claims** - 2+ independent sources
 4. **Note conflicting info** - Don't hide contradictions
 5. **Distinguish fact vs opinion** - Label speculation clearly
-6. **Reference sources** - Add source references near claims where applicable
+6. **CRITICAL: Always include inline citations** - Use `[citation:Title](URL)` format immediately after each claim from external sources
-7. **Update as you go** - Don't wait until end to synthesize
+7. **Extract URLs from search results** - web_search returns {title, url, snippet} - always use the URL field
+8. **Update as you go** - Don't wait until end to synthesize
+
+### Citation Examples
+
+**Good - With inline citations:**
+```markdown
+The project gained 10,000 stars within 3 months of launch [citation:GitHub Stats](https://github.com/owner/repo).
+The architecture uses LangGraph for workflow orchestration [citation:LangGraph Docs](https://langchain.com/langgraph).
+```
+
+**Bad - Without citations:**
+```markdown
+The project gained 10,000 stars within 3 months of launch.
+The architecture uses LangGraph for workflow orchestration.
+```

skills/public/github-deep-research/assets/report_template.md

@@ -30,6 +30,9 @@
 
 {EXECUTIVE_SUMMARY}
 
+**IMPORTANT**: Include inline citations using `[citation:Title](URL)` format after each claim. Example:
+"The project gained 10k stars in 3 months [citation:GitHub Stats](https://github.com/owner/repo)."
+
 ---
 
 ## Complete Chronological Timeline
@@ -56,6 +59,8 @@
 
 ## Key Analysis
 
+**IMPORTANT**: Support each analysis point with inline citations `[citation:Title](URL)`.
+
 ### {ANALYSIS_SECTION_1_TITLE}
 
 {ANALYSIS_SECTION_1_CONTENT}