Merge pull request #253 from khoj-ai/features/github-issues-indexing

Support indexing Github issues as well as corresponding comments

config/khoj_docker.yml

@@ -11,7 +11,7 @@ content-type:
 
   markdown:
     input-files: null
-    input-filter: ["/data/markdown/**/*.md"]
+    input-filter: ["/data/markdown/**/*.markdown"]
     compressed-jsonl: "/data/embeddings/markdown.jsonl.gz"
     embeddings-file: "/data/embeddings/markdown_embeddings.pt"
 

src/khoj/interface/web/index.html

@@ -24,23 +24,26 @@
         }
 
         function render_org(query, data, classPrefix="") {
-            var orgCode = data.map(function (item) {
-                return `${item.entry}`
-            }).join("\n")
-            var orgParser = new Org.Parser();
-            var orgDocument = orgParser.parse(orgCode);
-            var orgHTMLDocument = orgDocument.convert(Org.ConverterHTML, { htmlClassPrefix: classPrefix });
-            return `<div class="results-org">` + orgHTMLDocument.toString() + `</div>`;
+            return data.map(function (item) {
+                var orgParser = new Org.Parser();
+                var orgDocument = orgParser.parse(item.entry);
+                var orgHTMLDocument = orgDocument.convert(Org.ConverterHTML, { htmlClassPrefix: classPrefix });
+                return `<div class="results-org">` + orgHTMLDocument.toString() + `</div>`;
+            }).join("\n");
         }
 
         function render_markdown(query, data) {
             var md = window.markdownit();
             return data.map(function (item) {
+                let rendered = "";
                 if (item.additional.file.startsWith("http")) {
                     lines = item.entry.split("\n");
-                    return md.render(`${lines[0]}\t[*](${item.additional.file})\n${lines.slice(1).join("\n")}`);
+                    rendered = md.render(`${lines[0]}\t[*](${item.additional.file})\n${lines.slice(1).join("\n")}`);
                 }
-                return `<div class="results-markdown">` + md.render(`${item.entry}`) + `</div>`;
+                else {
+                    rendered = md.render(`${item.entry}`);
+                }
+                return `<div class="results-markdown">` + rendered + `</div>`;
             }).join("\n");
         }
 
@@ -59,16 +62,21 @@
             }).join("\n");
         }
 
-        function render_mutliple(query, data, type) {
+        function render_multiple(query, data, type) {
             let html = "";
             data.forEach(item => {
-               if (item.additional.file.endsWith(".org")) {
-                html += render_org(query, [item], "org-");
-               } else if (item.additional.file.endsWith(".md")) {
-                html += render_markdown(query, [item]);
-               } else if (item.additional.file.endsWith(".pdf")) {
-                html += render_pdf(query, [item]);
-               }
+                if (item.additional.file.endsWith(".org")) {
+                    html += render_org(query, [item], "org-");
+                } else if (
+                    item.additional.file.endsWith(".md") ||
+                    item.additional.file.endsWith(".markdown") ||
+                        (item.additional.file.includes("issues") && item.additional.file.includes("github.com"))
+                    )
+                {
+                    html += render_markdown(query, [item]);
+                } else if (item.additional.file.endsWith(".pdf")) {
+                    html += render_pdf(query, [item]);
+                }
             });
             return html;
         }
@@ -88,11 +96,25 @@
             } else if (type === "pdf") {
                 results = render_pdf(query, data);
             } else if (type === "github" || type === "all") {
-                results = render_mutliple(query, data, type);
+                results = render_multiple(query, data, type);
             } else {
                 results = data.map((item) => `<div class="results-plugin">` + `<p>${item.entry}</p>` + `</div>`).join("\n")
             }
-            return `<div id="results-${type}">${results}</div>`;
+
+            // Any POST rendering goes here.
+
+            let renderedResults = document.createElement("div");
+            renderedResults.id = `results-${type}`;
+            renderedResults.innerHTML = results;
+
+            // For all elements that are of type img in the results html and have a src with 'avatar' in the URL, add the class 'avatar'
+            // This is used to make the avatar images round
+            let images = renderedResults.querySelectorAll("img[src*='avatar']");
+            for (let i = 0; i < images.length; i++) {
+                images[i].classList.add("avatar");
+            }
+
+            return renderedResults.outerHTML;
         }
 
         function search(rerank=false) {
@@ -265,7 +287,6 @@
             margin: 0px;
             background: #f8fafc;
             color: #475569;
-            text-align: center;
             font-family: roboto, karma, segoe ui, sans-serif;
             font-size: 20px;
             font-weight: 300;
@@ -371,6 +392,28 @@
             max-width: 100;
         }
 
+        a {
+            color: #3b82f6;
+            text-decoration: none;
+        }
+
+        img.avatar {
+            width: 20px;
+            height: 20px;
+            border-radius: 50%;
+        }
+
+        div.results-markdown,
+        div.results-org,
+        div.results-pdf {
+            text-align: left;
+            box-shadow: 2px 2px 2px var(--primary-hover);
+            border-radius: 5px;
+            padding: 10px;
+            margin: 10px 0;
+            border: 1px solid rgb(229, 229, 229);
+        }
+
     </style>
 
 </html>

src/khoj/processor/github/github_to_jsonl.py

@@ -1,7 +1,8 @@
 # Standard Packages
 import logging
 import time
-from typing import Dict, List
+from datetime import datetime
+from typing import Dict, List, Union
 
 # External Packages
 import requests
@@ -13,6 +14,7 @@ from khoj.processor.markdown.markdown_to_jsonl import MarkdownToJsonl
 from khoj.processor.org_mode.org_to_jsonl import OrgToJsonl
 from khoj.processor.text_to_jsonl import TextToJsonl
 from khoj.utils.jsonl import dump_jsonl, compress_jsonl_data
+from khoj.utils.rawconfig import Entry
 
 
 logger = logging.getLogger(__name__)
@@ -22,6 +24,8 @@ class GithubToJsonl(TextToJsonl):
     def __init__(self, config: GithubContentConfig):
         super().__init__(config)
         self.config = config
+        self.session = requests.Session()
+        self.session.headers.update({"Authorization": f"token {self.config.pat_token}"})
 
     @staticmethod
     def wait_for_rate_limit_reset(response, func, *args, **kwargs):
@@ -53,6 +57,7 @@ class GithubToJsonl(TextToJsonl):
 
         logger.info(f"Found {len(markdown_files)} markdown files in github repo {repo_shorthand}")
         logger.info(f"Found {len(org_files)} org files in github repo {repo_shorthand}")
+        current_entries = []
 
         with timer(f"Extract markdown entries from github repo {repo_shorthand}", logger):
             current_entries = MarkdownToJsonl.convert_markdown_entries_to_maps(
@@ -65,6 +70,12 @@ class GithubToJsonl(TextToJsonl):
         with timer(f"Extract commit messages from github repo {repo_shorthand}", logger):
             current_entries += self.convert_commits_to_entries(self.get_commits(repo_url), repo)
 
+        with timer(f"Extract issues from github repo {repo_shorthand}", logger):
+            issue_entries = GithubToJsonl.convert_issues_to_entries(
+                *GithubToJsonl.extract_github_issues(self.get_issues(repo_url))
+            )
+            current_entries += issue_entries
+
         with timer(f"Split entries by max token size supported by model {repo_shorthand}", logger):
             current_entries = TextToJsonl.split_entries_by_max_tokens(current_entries, max_tokens=256)
 
@@ -102,7 +113,7 @@ class GithubToJsonl(TextToJsonl):
         contents = response.json()
 
         # Wait for rate limit reset if needed
-        result = self.wait_for_rate_limit_reset(response, self.get_files)
+        result = self.wait_for_rate_limit_reset(response, self.get_files, repo_url, repo)
         if result is not None:
             return result
 
@@ -130,35 +141,43 @@ class GithubToJsonl(TextToJsonl):
 
     def get_file_contents(self, file_url):
         # Get text from each markdown file
-        headers = {"Authorization": f"token {self.config.pat_token}", "Accept": "application/vnd.github.v3.raw"}
-        response = requests.get(file_url, headers=headers)
+        headers = {"Accept": "application/vnd.github.v3.raw"}
+        response = self.session.get(file_url, headers=headers, stream=True)
 
         # Wait for rate limit reset if needed
         result = self.wait_for_rate_limit_reset(response, self.get_file_contents, file_url)
         if result is not None:
             return result
 
-        return response.content.decode("utf-8")
+        content = ""
+        for chunk in response.iter_content(chunk_size=2048):
+            if chunk:
+                content += chunk.decode("utf-8")
+
+        return content
 
     def get_commits(self, repo_url: str) -> List[Dict]:
+        return self._get_commits(f"{repo_url}/commits")
+
+    def _get_commits(self, commits_url: Union[str, None]) -> List[Dict]:
         # Get commit messages from the repository using the Github API
-        commits_url = f"{repo_url}/commits"
-        headers = {"Authorization": f"token {self.config.pat_token}"}
         params = {"per_page": 100}
         commits = []
 
         while commits_url is not None:
             # Get the next page of commits
-            response = requests.get(commits_url, headers=headers, params=params)
-            raw_commits = response.json()
+            response = self.session.get(commits_url, params=params, stream=True)
+
+            # Read the streamed response into a JSON object
+            content = response.json()
 
             # Wait for rate limit reset if needed
-            result = self.wait_for_rate_limit_reset(response, self.get_commits)
+            result = self.wait_for_rate_limit_reset(response, self._get_commits, commits_url)
             if result is not None:
                 return result
 
             # Extract commit messages from the response
-            for commit in raw_commits:
+            for commit in content:
                 commits += [{"content": commit["commit"]["message"], "path": commit["html_url"]}]
 
             # Get the URL for the next page of commits, if any
@@ -166,6 +185,74 @@ class GithubToJsonl(TextToJsonl):
 
         return commits
 
+    def get_issues(self, repo_url: str) -> List[Dict]:
+        return self._get_issues(f"{repo_url}/issues")
+
+    def _get_issues(self, issues_url: Union[str, None]) -> List[Dict]:
+        issues = []
+        per_page = 30
+        params = {"per_page": per_page, "state": "all"}
+
+        while issues_url is not None:
+            # Get the next page of issues
+            response = self.session.get(issues_url, params=params, stream=True)  # type: ignore
+            raw_issues = response.json()
+
+            # Wait for rate limit reset if needed
+            result = self.wait_for_rate_limit_reset(response, self._get_issues, issues_url)
+            if result is not None:
+                return result
+
+            for issue in raw_issues:
+                username = issue["user"]["login"]
+                user_url = f"[{username}]({issue['user']['html_url']})"
+                issue_content = {
+                    "content": f"## [Issue {issue['number']}]({issue['html_url']}) {issue['title']}\nby {user_url}\n\n{issue['body']}",
+                    "path": issue["html_url"],
+                }
+                issue_content["created_at"] = {issue["created_at"]}
+                if issue["comments"] > 0:
+                    issue_content["comments"] = self.get_comments(issue["comments_url"])
+                issues += [issue_content]
+
+            issues_url = response.links.get("next", {}).get("url")
+
+        return issues
+
+    def get_comments(self, comments_url: Union[str, None]) -> List[Dict]:
+        # By default, the number of results per page is 30. We'll keep it as-is for now.
+        comments = []
+        per_page = 30
+        params = {"per_page": per_page}
+
+        while comments_url is not None:
+            # Get the next page of comments
+            response = self.session.get(comments_url, params=params, stream=True)
+            raw_comments = response.json()
+
+            # Wait for rate limit reset if needed
+            result = self.wait_for_rate_limit_reset(response, self.get_comments, comments_url)
+            if result is not None:
+                return result
+
+            for comment in raw_comments:
+                created_at = datetime.strptime(comment["created_at"], "%Y-%m-%dT%H:%M:%SZ").strftime("%Y-%m-%d %H:%M")
+                commenter = comment["user"]["login"]
+                commenter_url = comment["user"]["html_url"]
+                comment_url = comment["html_url"]
+                comment_url_link = f"[{created_at}]({comment_url})"
+                avatar_url = comment["user"]["avatar_url"]
+                avatar = f"![{commenter}]({avatar_url})"
+                comments += [
+                    {
+                        "content": f"### {avatar} [{commenter}]({commenter_url}) - ({comment_url_link})\n\n{comment['body']}"
+                    }
+                ]
+
+            comments_url = response.links.get("next", {}).get("url")
+
+        return comments
+
     def convert_commits_to_entries(self, commits, repo: GithubRepoConfig) -> List[Entry]:
         entries: List[Entry] = []
         for commit in commits:
@@ -201,3 +288,32 @@ class GithubToJsonl(TextToJsonl):
                 doc["content"], doc["path"], entries, entry_to_file_map
             )
         return entries, dict(entry_to_file_map)
+
+    @staticmethod
+    def extract_github_issues(issues):
+        entries = []
+        entry_to_file_map = {}
+        for issue in issues:
+            content = issue["content"]
+            if "comments" in issue:
+                for comment in issue["comments"]:
+                    content += "\n\n" + comment["content"]
+            entries.append(content)
+            entry_to_file_map[content] = {"path": issue["path"]}
+        return entries, entry_to_file_map
+
+    @staticmethod
+    def convert_issues_to_entries(parsed_entries: List[str], entry_to_metadata_map: Dict[str, Dict]) -> List[Entry]:
+        entries = []
+        for entry in parsed_entries:
+            entry_file_name = entry_to_metadata_map[entry]["path"]
+            entries.append(
+                Entry(
+                    compiled=entry,
+                    raw=entry,
+                    heading=entry.split("\n")[0],
+                    file=entry_file_name,
+                )
+            )
+
+        return entries

src/khoj/routers/api.py

@@ -216,6 +216,20 @@ async def search(
                 )
             ]
 
+        if (t == SearchType.Github or t == SearchType.All) and state.model.github_search:
+            # query github issues
+            search_futures += [
+                executor.submit(
+                    text_search.query,
+                    user_query,
+                    state.model.github_search,
+                    question_embedding=encoded_asymmetric_query,
+                    rank_results=r or False,
+                    score_threshold=score_threshold,
+                    dedupe=dedupe or True,
+                )
+            ]
+
         if (t == SearchType.Pdf or t == SearchType.All) and state.model.pdf_search:
             # query pdf files
             search_futures += [

tests/conftest.py

@@ -120,7 +120,7 @@ def md_content_config(tmp_path_factory):
     content_config = ContentConfig()
     content_config.markdown = TextContentConfig(
         input_files=None,
-        input_filter=["tests/data/markdown/*.md"],
+        input_filter=["tests/data/markdown/*.markdown"],
         compressed_jsonl=content_dir.joinpath("markdown.jsonl"),
         embeddings_file=content_dir.joinpath("markdown_embeddings.pt"),
     )