Skip to content

first commit #106

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
tejjgv wants to merge 1 commit into
base: main
Choose a base branch
from
Open

first commit #106

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions Apply-agent
Open in desktop
Submodule Apply-agent added at 4a8d52
161 changes: 123 additions & 38 deletions src/applypilot/apply/launcher.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -110,10 +110,30 @@ def acquire_job(target_url: str | None = None, min_score: int = 7,
fit_score, location, full_description, cover_letter_path
FROM jobs
WHERE (url = ? OR application_url = ? OR application_url LIKE ? OR url LIKE ?)
AND tailored_resume_path IS NOT NULL
AND apply_status != 'in_progress'
AND (apply_status IS NULL OR apply_status != 'in_progress')
LIMIT 1
""", (target_url, target_url, like, like)).fetchone()

if not row:
from urllib.parse import urlparse
parsed = urlparse(target_url)
site = parsed.netloc.replace("www.", "") or "Custom"
path_parts = [p for p in parsed.path.split('/') if p]
title = path_parts[-1].replace('-', ' ').replace('_', ' ').title() if path_parts else "Job Application"
now = datetime.now(timezone.utc).isoformat()

conn.execute("""
INSERT OR IGNORE INTO jobs (url, application_url, title, site, fit_score, strategy, discovered_at)
VALUES (?, ?, ?, ?, 10, 'manual_url', ?)
""", (target_url, target_url, title, site, now))
conn.commit()

row = conn.execute("""
SELECT url, title, site, application_url, tailored_resume_path,
fit_score, location, full_description, cover_letter_path
FROM jobs
WHERE url = ?
""", (target_url,)).fetchone()
else:
blocked_sites, blocked_patterns = _load_blocked()
# Build parameterized filters to avoid SQL injection
Expand All @@ -131,13 +151,12 @@ def acquire_job(target_url: str | None = None, min_score: int = 7,
SELECT url, title, site, application_url, tailored_resume_path,
fit_score, location, full_description, cover_letter_path
FROM jobs
WHERE tailored_resume_path IS NOT NULL
AND (apply_status IS NULL OR apply_status = 'failed')
WHERE (apply_status IS NULL OR apply_status = 'failed')
AND (apply_attempts IS NULL OR apply_attempts < ?)
AND fit_score >= ?
{site_clause}
{url_clauses}
ORDER BY fit_score DESC, url
ORDER BY (CASE WHEN tailored_resume_path IS NOT NULL THEN 0 ELSE 1 END), fit_score DESC, url
LIMIT 1
""", [config.DEFAULTS["max_apply_attempts"]] + params).fetchone()

Expand Down Expand Up @@ -221,12 +240,14 @@ def gen_prompt(target_url: str, min_score: int = 7,
if not job:
return None

# Read resume text
# Read resume text (tailored or main master resume fallback)
resume_path = job.get("tailored_resume_path")
txt_path = Path(resume_path).with_suffix(".txt") if resume_path else None
txt_path = Path(resume_path).with_suffix(".txt") if resume_path and Path(resume_path).exists() else None
resume_text = ""
if txt_path and txt_path.exists():
resume_text = txt_path.read_text(encoding="utf-8")
elif config.RESUME_PATH.exists():
resume_text = config.RESUME_PATH.read_text(encoding="utf-8")

prompt = prompt_mod.build_prompt(job=job, tailored_resume=resume_text)

Expand Down Expand Up @@ -303,12 +324,14 @@ def run_job(job: dict, port: int, worker_id: int = 0,
'applied', 'expired', 'captcha', 'login_issue',
'failed:reason', or 'skipped'.
"""
# Read tailored resume text
# Read tailored resume text (or main master resume fallback)
resume_path = job.get("tailored_resume_path")
txt_path = Path(resume_path).with_suffix(".txt") if resume_path else None
txt_path = Path(resume_path).with_suffix(".txt") if resume_path and Path(resume_path).exists() else None
resume_text = ""
if txt_path and txt_path.exists():
resume_text = txt_path.read_text(encoding="utf-8")
elif config.RESUME_PATH.exists():
resume_text = config.RESUME_PATH.read_text(encoding="utf-8")

# Build the prompt
agent_prompt = prompt_mod.build_prompt(
Expand All @@ -317,48 +340,85 @@ def run_job(job: dict, port: int, worker_id: int = 0,
dry_run=dry_run,
)

# Write per-worker MCP config
# Detect configured agent CLI
agent_info = config.get_agent_cli_info()
if not agent_info:
raise RuntimeError("No browser agent CLI found. Please install Claude Code CLI or Antigravity CLI (agy).")

agent_type, agent_exec = agent_info

# Worker directory
worker_dir = reset_worker_dir(worker_id)
mcp_data = _make_mcp_config(port)

# Write per-worker MCP configs
mcp_config_path = config.APP_DIR / f".mcp-apply-{worker_id}.json"
mcp_config_path.write_text(json.dumps(_make_mcp_config(port)), encoding="utf-8")

# Build claude command
cmd = [
"claude",
"--model", model,
"-p",
"--mcp-config", str(mcp_config_path),
"--permission-mode", "bypassPermissions",
"--no-session-persistence",
"--disallowedTools", (
"mcp__gmail__draft_email,mcp__gmail__modify_email,"
"mcp__gmail__delete_email,mcp__gmail__download_attachment,"
"mcp__gmail__batch_modify_emails,mcp__gmail__batch_delete_emails,"
"mcp__gmail__create_label,mcp__gmail__update_label,"
"mcp__gmail__delete_label,mcp__gmail__get_or_create_label,"
"mcp__gmail__list_email_labels,mcp__gmail__create_filter,"
"mcp__gmail__list_filters,mcp__gmail__get_filter,"
"mcp__gmail__delete_filter"
),
"--output-format", "stream-json",
"--verbose", "-",
]
mcp_config_path.write_text(json.dumps(mcp_data), encoding="utf-8")
# Also write mcp_config.json into worker_dir for Antigravity (agy)
(worker_dir / "mcp_config.json").write_text(json.dumps(mcp_data), encoding="utf-8")

# For agy: update the global playwright MCP server to point at the correct CDP port
if agent_type == "antigravity":
try:
subprocess.run(
[
agent_exec, "mcp", "add",
"playwright",
"npx", "@playwright/mcp@latest",
f"--cdp-endpoint=http://localhost:{port}",
f"--viewport-size={config.DEFAULTS['viewport']}",
],
check=False, capture_output=True, timeout=10,
)
except Exception as e:
logger.warning("Could not update agy playwright MCP config: %s", e)

# Build agent command
if agent_type == "antigravity":
cmd = [
agent_exec,
"--dangerously-skip-permissions",
"--output-format", "stream-json",
"-p", agent_prompt,
]
if model and model not in ("sonnet", "haiku"):
cmd.extend(["--model", model])
else:
cmd = [
agent_exec,
"--model", model,
"-p",
"--mcp-config", str(mcp_config_path),
"--permission-mode", "bypassPermissions",
"--no-session-persistence",
"--disallowedTools", (
"mcp__gmail__draft_email,mcp__gmail__modify_email,"
"mcp__gmail__delete_email,mcp__gmail__download_attachment,"
"mcp__gmail__batch_modify_emails,mcp__gmail__batch_delete_emails,"
"mcp__gmail__create_label,mcp__gmail__update_label,"
"mcp__gmail__delete_label,mcp__gmail__get_or_create_label,"
"mcp__gmail__list_email_labels,mcp__gmail__create_filter,"
"mcp__gmail__list_filters,mcp__gmail__get_filter,"
"mcp__gmail__delete_filter"
),
"--output-format", "stream-json",
"--verbose", "-",
]

env = os.environ.copy()
env.pop("CLAUDECODE", None)
env.pop("CLAUDE_CODE_ENTRYPOINT", None)

worker_dir = reset_worker_dir(worker_id)

update_state(worker_id, status="applying", job_title=job["title"],
company=job.get("site", ""), score=job.get("fit_score", 0),
start_time=time.time(), actions=0, last_action="starting")
add_event(f"[W{worker_id}] Starting: {job['title'][:40]} @ {job.get('site', '')}")
add_event(f"[W{worker_id}] Starting [{agent_type}]: {job['title'][:40]} @ {job.get('site', '')}")

worker_log = config.LOG_DIR / f"worker-{worker_id}.log"
ts_header = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
log_header = (
f"\n{'=' * 60}\n"
f"[{ts_header}] {job['title']} @ {job.get('site', '')}\n"
f"[{ts_header}] [{agent_type}] {job['title']} @ {job.get('site', '')}\n"
f"URL: {job.get('application_url') or job['url']}\n"
f"Score: {job.get('fit_score', 'N/A')}/10\n"
f"{'=' * 60}\n"
Expand All @@ -383,7 +443,8 @@ def run_job(job: dict, port: int, worker_id: int = 0,
with _claude_lock:
_claude_procs[worker_id] = proc

proc.stdin.write(agent_prompt)
if agent_type == "claude":
proc.stdin.write(agent_prompt)
proc.stdin.close()

text_parts: list[str] = []
Expand All @@ -397,6 +458,9 @@ def run_job(job: dict, port: int, worker_id: int = 0,
try:
msg = json.loads(line)
msg_type = msg.get("type")
event = msg.get("event")

# Claude Code JSON stream format
if msg_type == "assistant":
for block in msg.get("message", {}).get("content", []):
bt = block.get("type")
Expand Down Expand Up @@ -437,6 +501,27 @@ def run_job(job: dict, port: int, worker_id: int = 0,
"turns": msg.get("num_turns", 0),
}
text_parts.append(msg.get("result", ""))

# Antigravity (agy) JSON stream format
elif event == "step_update":
su = msg.get("step_update", {})
text_delta = su.get("text_delta")
if text_delta:
text_parts.append(text_delta)
lf.write(text_delta)
elif event == "result":
res = msg.get("result", {})
if isinstance(res, dict):
usage = res.get("usage", {})
stats = {
"input_tokens": usage.get("input_tokens", 0),
"output_tokens": usage.get("output_tokens", 0),
"cache_read": usage.get("cache_read_tokens", 0),
"turns": res.get("num_turns", 0),
}
resp_text = res.get("response", "")
text_parts.append(resp_text)
lf.write(resp_text + "\n")
except json.JSONDecodeError:
text_parts.append(line)
lf.write(line + "\n")
Expand Down
34 changes: 27 additions & 7 deletions src/applypilot/apply/prompt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -440,13 +440,33 @@ def build_prompt(job: dict, tailored_resume: str,
personal = profile["personal"]

# --- Resolve resume PDF path ---
resume_path = job.get("tailored_resume_path")
if not resume_path:
raise ValueError(f"No tailored resume for job: {job.get('title', 'unknown')}")

src_pdf = Path(resume_path).with_suffix(".pdf").resolve()
if not src_pdf.exists():
raise ValueError(f"Resume PDF not found: {src_pdf}")
src_pdf: Path | None = None
tailored_path = job.get("tailored_resume_path")
if tailored_path:
cand = Path(tailored_path).with_suffix(".pdf").resolve()
if cand.exists():
src_pdf = cand

if src_pdf is None:
if config.RESUME_PDF_PATH.exists():
src_pdf = config.RESUME_PDF_PATH.resolve()
elif config.RESUME_PATH.exists():
cand_pdf = config.RESUME_PATH.with_suffix(".pdf").resolve()
if cand_pdf.exists():
src_pdf = cand_pdf
else:
try:
from applypilot.scoring.pdf import convert_to_pdf
src_pdf = convert_to_pdf(config.RESUME_PATH, config.RESUME_PDF_PATH).resolve()
except Exception as e:
logger.warning("Could not convert main resume to PDF: %s", e)

if not src_pdf or not src_pdf.exists():
raise ValueError(f"No resume (tailored or main master) found for job: {job.get('title', 'unknown')}")

# Fall back to main master resume text if tailored_resume string is empty
if not tailored_resume and config.RESUME_PATH.exists():
tailored_resume = config.RESUME_PATH.read_text(encoding="utf-8")

# Copy to a clean filename for upload (recruiters see the filename)
full_name = personal["full_name"]
Expand Down
Loading